Computer graphic display visualization system and method

Abstract

An improved human user computer interface system, providing a graphic representation of a hierarchy populated with naturally classified objects, having included therein at least one associated object having a distinct classification. Preferably, a collaborative filter is employed to define the appropriate associated object. The associated object preferably comprises a sponsored object, generating a subsidy or revenue.

Claims

What I claim is:

1. A method of providing a human-computer user interface, comprising the steps of:

(a) receiving an input through a user interface providing the user with navigational tools for defining and retrieving objects based on a resource locator thereof;

(b) providing an object search engine for selecting a set of objects according to a user-defined content criteria from a larger set of objects including objects of varying relevance to the user-defined content criteria accessed through the user interface and returning respective resource locators of selected objects, the object search engine employing at least first and second algorithms for selecting respectively different portions of the set of objects;

(c) providing a hierarchal organizational structure for the set of selected respective resource locators of selected objects, having at least two resource locators for objects organized within a single hierarchal level, for presentation to the user through the user interface, wherein resource locators for objects selected according to the first algorithm are automatically organized within the hierarchal organizational structure based on an associated object content, and resource locators for objects selected according to the second algorithm are automatically organized within the hierarchal organizational structure based on at least one criterion independent of an associated object content.

2. The method according to claim 1, further comprising the step of inserting objects extrinsic to the user-defined content criteria into the hierarchal organizational structure of selected objects.

3. The method according to claim 1, wherein the first algorithm selects objects based on a relevance of an associated content with the user-defined content criteria and the second algorithm selects an advertisement object selected based on the user-defined content criteria.

4. The method according to claim 3, wherein a commercial sponsor pays for advertisement objects based on a semantic context of message delivery.

5. The method according to claim 3, wherein a commercial sponsor pays for delivery of advertisement objects based on a value of a subsequent commercial transaction with the user.

6. The method according to claim 3, wherein the advertisement objects are identified through a collaborative filter process.

7. The method according to claim 3, wherein the advertisement objects are contextually related to the user-defined content criteria.

8. The method according to claim 1, wherein the first algorithm selects objects based on a relevance of an associated content with the user-defined content criteria and the second algorithm selects objects identified through a collaborative filter process.

9. The method according to claim 1, wherein the first algorithm selects objects based on a relevance of an associated content with the user-defined content criteria and the second algorithm selects objects which are contextually related to the user-defined content criteria.

10. The method according to claim 1, wherein the objects selected by both the first and second d algorithms are contextually appropriate for a respective positioning within the hierarchal organizational structure.

11. The method according to claim 1, wherein the hierarchal organizational structure comprises a tree structure displaying at least three hierarchal levels.

12. The method according to claim 1, wherein the hierarchal organizational structure comprises a hyperbolic tree structure.

13. The method according to claim 1, wherein the hierarchal organizational structure comprises a display generated by a hyperbolic tree applet.

14. The method according to claim 1, wherein the hierarchal organizational structure comprises a state independent information object.

15. The method according to claim 1, further comprising the step of ranking members of the set of selected objects based on a correspondence to the user-defined content criteria.

16. The method according to claim 1, further comprising the step of receiving a ranking preference from the user for a ranking method for ranking members of the set of selected objects.

17. The method according to claim 1, further comprising the step of representing a history of access to the set of objects.

18. The method according to claim 1, fisher comprising the steps of manipulating an object within the hierarchal organizational structure through the graphic user interface, and requesting information content associated with the manipulated object.

19. The method according to claim 1, wherein at least two distinct predetermined hierarchical organizations of information are provided, each having at least three hierarchal levels for a universe of objects, further comprising the steps of:

(a) defining a relevant hierarchy from among the at least two distinct predetermined hierarchical organizations of information;

(d) displaying links to the set of objects according to the relevant hierarchy; and

(e) storing at least a subset of the presented links within the relevant hierarchy as a state independent object.

20. The method according to claim 1, further comprising the step of defining a user profile, for modifying the selection by the object search engine, and wherein user profile is stored in an encrypted form which is resistant to detailed interrogation.

21. The method according to claim 1, further comprising the step of presenting the hierarchal organizational structure with an applet, wherein the returned respective resource locators of selected objects are transmitted to the applet, which formats the set of selected objects in a graphic format hierarchal organizational structure, based on a relationship of a content corresponding to each object.

22. The method according to claim 1, further comprising the step of providing an adaptive user profile applet, comprising a collaborative filter for initial classification, which subsequently is modified based on user observation, wherein the user-defined content criteria is based on an explicit user input and a function of the adaptive user profile applet.

23. The method according to claim 1, further comprising the step of defining the hierarchal organizational structure as a user taxonomic hierarchy of interests, correlating the user taxonomic hierarchy with a set of references taxonomic hierarchies, and modifying the user taxonomic hierarchy based on sets of rules associated with a reference taxonomic hierarchies having high correlations.

24. The method according to claim 1, wherein at least one object has an associated digital rights rule, farther comprising the step of applying digital rights rules to accesses of objects by the user.

25. The method according to claim 24, wherein at least one digital rights rule provides a positive incentive to the user.

26. A computer readable medium having stored thereon a software program for executing the method according to claim 1.

27. The method according to claim 1, wherein objects selected according to the at least first and second algorithms are differentiated within the hierarchal organizational structure.

28. The method according to claim 1, wherein at least one algorithm for selecting objects operates to generate a commercial subsidy for use of the object search engine.

29. The method according to claim 1, wherein at least one algorithm for selecting objects operates to generate a commercial subsidy for use of the object search engine.

30. The method according to claim 1, further comprising the step of ranking members of the set of selected objects independent of the user-defined content criteria.

31. The method according to claim 1, wherein

(a) the user-defined content criteria comprises a semantic query,

(b) the hierarchal organizational structure is taxonomic, objects selected according to the first algorithm being organized within the hierarchy based on a semantic classification of the respective object content, objects selected according to the second algorithm being organized based on a semantic relation of the selected object to the semantic query.

32. The method according to claim 1, further comprising the steps of producing a ranking for at least a portion of the selected objects within a single hierarchal level; and presenting the selected objects within the hierarchal organization structure in dependence on the ranking.

33. A system for providing a human-computer user interface, comprising:

(a) a set of navigational tools for defining and retrieving objects based on a resource locator thereof;

(b) an object search engine for selecting a set of objects according to a user-defined content criterion and returning respective resource locators of selected objects, the object search engine employing at least first and second schemes for selecting objects; and

(c) means for presenting a hierarchal organizational structure for the set of selected objects, wherein at least one level of the hierarchal organizational structure has at least two objects organized therein, and wherein at least a portion of the selected objects are organized within the hierarchal organizational structure based on an associated content and a respective scheme employed to select that object, the hierarchal organizational structure further including at least one object extrinsic to the selected objects.

34. The system according to claim 33, wherein objects extrinsic to the user-defined search criteria are inserted into the hierarchal organizational structure of selected objects based on a semantic relationship to at least one of the search criteria and selected objects.

35. The system according to claim 33, wherein the extrinsic objects comprise commercial messages.

36. The method according to claim 33, wherein the extrinsic objects comprise objects are identified through a collaborative filter process.

37. The system according to claim 33, wherein the extrinsic objects are contextually related to the user-defined content criteria.

38. A method of visualization of a set of objects, comprising:

(a) defining a taxonomic hierarchy, each hierarchal level within the hierarchy, below a top level, having at least one object, the at least one object having one parent hierarchal object and optionally a set of child objects, with a set of content objects populating the hierarchy, wherein at least one level of the hierarchy has at least two objects;

(b) defining, based on a user input, a selected object within the hierarchy for examination; and

(c) generating a display presenting the selected object and any child objects thereof; a representation of parental objects within the hierarchy, wherein each of the parent and child objects is associated with a hyperlink, a selection of a respective hyperlink serving to transform that object into the selected element, wherein when an object representing information content is selected, an associated set of related objects extrinsic to the defined hierarchy of objects and related to the taxonomic hierarchal level is displayed.

39. The method according to claim 38, wherein the associated set of related objects is defined by a process of collaborative filtering.

40. The method according to claim 38, wherein the content object defines a product promoted for sale.

41. A method of visualization of a set of elements, comprising:

(a) defining a natural hierarchy of objects wherein at least one level of the hierarchy has at least two objects;

(b) receiving a user limiter to define a set of objects in the hierarchy having natural hierarchal relationships;

(c) inserting at least one object extrinsic to the user limiter within the hierarchy of objects to provide artificial hierarchal relationships;

(d) displaying the set of objects and extrinsic objects with a graphic representation of the natural and artificial hierarchal relationships.

42. The method according to claim 41, wherein the inserting is controlled by a process of collaborative filtering.

43. The method according to claim 41, wherein the inserting is based on an advertising payment.

44. A method of providing a human-computer user interface, comprising the steps of:

(a) providing an object browser;

(b) receiving a user-defined content-based selection criteria and returning respective resource locators of selected objects consistent with the criteria and respective resource locators of objects associated with the criteria; and

(c) displaying the respective resource locators of the selected objects through the object browser, within a hierarchy, wherein objects selected based on content are placed in the hierarchy in content-dependent manner, and objects selected based on association with the criteria are placed in the hierarchy in association-dependent manner, wherein at least one level of the hierarchy has at least two objects.

45. The method according to claim 44, wherein the hierarchy is adaptive to the set of selected objects.

46. The method according to claim 44, wherein the objects selected based on association with the criteria have a semantic relationship with at least one of the selection criteria and the objects selected based on content.

47. The method according to claim 44, wherein the at least one object outside the set of selected objects is associated with a subsidy.

48. A method, comprising the steps of:

(a) receiving an input from a user comprising a content selection criteria;

(b) selecting a set of objects in dependence on the content selection criteria;

(c) automatically populating a hierarchal organizational structure with the selected objects, in dependence on an associated selected object content; and

(d) additionally automatically populating the hierarchal organization structure with a set of additional objects selected independent of an associated selected object content, the additional objects being populated in dependence on a relation of a respective additional object and the input,

wherein the hierarchal organization structure has at least one level having at least two objects.

49. The method according to claim 48, wherein the set of additional objects comprises objects associated with a commercial subsidy.

50. The method according to claim 48, wherein the hierarchal organizational structure is predetermined.

51. The method according to claim 48, wherein the hierarchal organizational structure is defined in an ad hoc manner.

52. The method according to claim 48, wherein the hierarchal organizational structure is communicated electronically to a user.

53. The method according to claim 48, wherein the hierarchal organizational structure is communicated graphically to a user.

54. The method according to claim 48, wherein the hierarchal organizational structure is communicated in an interactive form to a user.

55. A system comprising:

(a) an input for receiving a content selection criteria from a user;

(b) at least one processor for selecting a set of objects in dependence on the content selection criteria and automatically populating a hierarchal organizational structure with the selected objects, in dependence on an associated selected object content, and additionally automatically populating the hierarchal organization structure with a set of additional objects selected independent of a respective additional object content, in dependence on the input, the hierarchal organization structure having at least one hierarchal level having at least two objects; and

(c) an output for communicating at least a portion of the populated hierarchal organization structure with the user.

56. The system according to claim 55, further comprising an accounting system for accounting for a selection of said additional objects.

57. The system according to claim 55, wherein the hierarchal organizational structure is predetermined.

58. The system according to claim 55, wherein the hierarchal organizational structure is defined in an ad hoc manner.

Description

FIELD OF THE INVENTION

The present invention relates to the field of human computer interface systems, and more particularly to the field of improved graphic user interfaces for information retrieval systems.

BACKGROUND OF THE INVENTION

For almost as long as computers have existed, their designers and users have sought improvements to the user interface. Especially as computing power has increased, a greater portion of the available processing capacity has been devoted to improved interface design. Recent examples have been Microsoft Windows variants and Internet web browsers. Graphic interfaces provide significant flexibility to present data using various paradigms, and modern examples support use of data objects and applets. Traditional human computer interfaces have emphasized uniformity and consistency, thus, experienced users had a shortened learning curve for use of software and systems; while novice users often required extensive instruction before profitable use of a system. More recently, intuitive, adaptable and adaptive software interfaces have been proposed, which potentially allow faster adoption of the system by new users but which requires continued attention by experienced users due to the possibility of interface transformation.

While many computer applications are used both on personal computers and networked systems, the field of information retrieval and database access for casual users has garnered considerable interest. The Internet presents a vast relatively unstructured repository for information, leading to a need for Internet search engines and access portals based on Internet navigation. At this time, the Internet is gaining popularity because of its "universal" access, low access and information distribution costs, and suitability for conducting commercial transactions. However, this popularity, in conjunction with the non-standardized methods of presenting data and fantastic growth rate, have made locating desired information and navigation through the vast space difficult. Thus, improvements in human consumer interfaces for relatively unstructured data sets are desirable, wherein subjective improvements and wholesale adoption of new paradigms may both be valuable, including improved methods for searching and navigating the Internet.

Generally speaking, search engines for the World Wide Web (WWW, or simply "Web") aid users in locating resources among the estimated present one billion addressable sites on the Web. Search engines for the web generally employ a type of computer software called a "spider" to scan a proprietary database that is a subset of the resources available on the Web. Major known commercial search engines include such names as Yahoo, Excite, and Infoseek. Also known in the field are "metasearch engines," such as Dogpile and Metasearch, which compile and summarize the results of other search engines without generally themselves controlling an underlying database or using their own spider. All the search engines and metasearch engines, which are servers, operate with the aid of a browser, which are clients, and deliver to the client a dynamically generated web page which includes a list of hyperlinked universal resource locators (URLS) for directly accessing the referenced documents themselves by the web browser.

A Uniform Resource Identifier (RFC 1630) is the name for the standard generic object in the World Wide Web. Internet space is inhabited by many points of content. A URI (Uniform Resource Identifier is the way you identify any of those points of content, whether it be a page of text, a video or sound clip, a still or animated image, or a program. The most common form of URI is the Web page address, which is a particular form or subset of URI called a Uniform Resource Locator (URL). A URI typically describes: the mechanism used to access the resource; the specific computer that the resource is housed in; and the specific name of the resource (a file name) on the computer. Another kind of URI is the Uniform Resource Name (URN). A URN is a form of URI that has "institutional persistence," which means that its exact location may change from time to time, but some agency will be able to find it.

The structure of the World Wide Web includes multiple servers at distinct nodes of the Internet, each of which hosts a web server which transmits a web page in hypertext markup language (HTML) or extensible markup language (XML) (or a similar scheme) using the hypertext transport protocol (http). Each web page may include embedded hypertext linkages, which direct the client browser to other web pages, which may be hosted within any server on the network. A domain name server translates a top-level domain (TLD) name into an Internet protocol (IP) address, which identifies the appropriate server. Thus, Internet web resources, which are typically the aforementioned web pages, are thus typically referenced with a URL, which provides the TLD or IP address of the server, as well a hierarchal address for defining a resource of the server, e.g., a directory path on a server system.

A hypermedia collection may be represented by a directed graph having nodes that represent resources and arcs that represent embedded links between resources. Typically, a user interface, such as a browser, is utilized to access hyperlinked information resources. The user interface displays information "pages" or segments and provides a mechanism by which that user may follow the embedded hyperlinks. Many user interfaces allow selection of hyperlinked information via a pointing device, such as a mouse. Once selected, the system retrieves the information resource corresponding to the embedded hyperlink.

One approach to assisting users in locating information of interest within a collection is to add structure to the collection. For example, information is often sorted and classified so that a large portion of the collection need not be searched. However, this type of structure often requires some familiarity with the classification system, to avoid elimination of relevant resources by improperly limiting the search to a particular classification or group of classifications. Another approach used to locate information of interest to a user, is to couple resources through cross-referencing. Conventional cross-referencing of publications using citations provides the user enough information to retrieve a related publication, such as the author, tide of publication, date of publication, and the like. However, the retrieval process is often time-consuming and cumbersome. A more convenient, automated method of cross-referencing related documents utilizes hypertext or hyperlinks. Hyperlink systems allow authors or editors to embed links within their resources to other portions of those resources or to related resources in one or more collections that may be locally accessed, or remotely accessed via a network. Users of hypermedia systems can then browse through the resources by following the various links embedded by the authors or editors. These systems greatly simplify the task of locating and retrieving the documents when compared to a traditional citation, since the hyperlink is usually transparent to the user. Once selected, the system utilizes the embedded hyperlink to retrieve the associated resource and present it to the user, typically in a matter of seconds. The retrieved resource may contain additional hyperlinks to other related information that can be retrieved in a similar manner.

A well-recognized problem with existing search engines is the tendency to return hits for a query that are so incredibly numerous, sometimes in the hundreds, thousands, or even millions, that it is impractical for user to wade through them and find relevant results. Many users, probably the majority, would say that the existing technology returns far too much "garbage" in relation to pertinent results. This has lead to the desire among many users for an improved search engine, and in particular an improved Internet search engine.

In response the garbage problem, search engines have sought to develop unique proprietary approaches to gauging the relevance of results in relation to a user's query. Such technologies employ algorithms for either limiting the records returned in the selection process (the search) and/or by sorting selected results from the database according to a rank or weighting, which maybe predetermined or computed on the fly. The known techniques include counting the frequency or proximity of keywords, measuring the frequency of user visits to a site or the persistence of users on that site, using human librarians to estimate the value of a site and to quantify or rank it, measuring the extent to which the site is linked to other sites through ties called "hyperlinks" (see, Google.com and Clever.com), measuring how much economic investment is going into a site (Thunderstone.com), taking polls of users, or even ranking relevance in certain cases according to advertiser's willingness to bid the highest price for good position within ranked lists. As a result of relevance testing procedures, many search engines return hits in presumed rank order or relevance, and some place a percentage next to each hit which is said to represent the probability that the hit is relevant to the query, with the hits arranged in descending percentage order.

However, despite the apparent sophistication of many of the relevance testing techniques employed, the results typically fall short of the promise. Thus, there remains a need for a search engine for uncontrolled databases that provides to the user results, which accurately correspond the desired information sought.

Therefore, the art requires improved searching strategies and tools to provide increased efficiency in locating a user's desired content, while preventing dilution of the best records with those that are redundant, off-topic or irrelevant, or directed to a different audience.

Commercial Subsidy (Advertising)

Advertisers are generally willing to pay more to deliver an impression (e.g., a banner ad or other type of advertisement) to users who are especially sensitive to advertisements for their products or are seeking to purchase products corresponding to those sold by the advertisers, and the economic model often provides greater compensation in the event of a "click through", which is a positive action taken by the user to interact with the ad to receive further information.

This principle, of course, actually operates correspondingly in traditional media. For example, a bicycle manufacturer in generally is willing to pay more per subscriber to place advertisements in a magazine having content directed to bicycle buffs than in a general interest publication. However, this principle has not operated very extensively in the search engine marketplace, partly because there is little differentiation among the known characteristics of the users of particular search engines, and because, even after a search inquiry in submitted, there may be little basis on which to judge what user's intention or interest really is, owing to the generality or ambiguity of user's request, so that even after a search request is processed, it may be impossible to estimate the salient economic, demographic, purchasing or interest characteristics of the user in the context of a particular search. In fact, some "cookie" based mechanisms provide long-term persistence of presumed characteristics even when these might be determined to be clearly erroneous. Thus, the existing techniques tend to exaggerate short term, ignorance based or antithetical interests of the user, since these represent the available data set. For example, if a child seeks to research the evils of cigar smoking for a school class project, a search engine might classify the user as a person interested in cigar smoking and cigar paraphernalia, which is clearly not the case. Further, the demographics of a cigar aficionado might tempt an advertiser of distilled liquors to solicit this person as a potential client. The presumed interest in cigars and liquor might then result in adult-oriented materials being presented. Clearly, the simple presumptions that are behind this parade of horribles may often result in erroneous conclusions.

Although a few search engines for the mass market exist that charge a fee for use, this model has not been popular or successful. Instead, most search engines offer free access, subject to user tolerating background advertising or pitches for electronic commerce sales or paid links to sites that offer goods and services, including the aforementioned banner ads. These advertisements are typically paid for by sponsors on a per impression basis (each time a user opens the page on which the banner ad appears) or on a "click-through basis" (normally a higher charge, because user has decided to select the ad and "open it up" by activating an underlying hyper-link). In addition, most search engines seek "partners" with whom they mutually share hyperlinks to each other's sites. Finally, the search engines may seek to offer shopping services or merchandise opportunities, and the engines may offer these either globally to all users, or on a context sensitive basis responsive to a user's particular search.

Targeted Advertising

The current wide-ranging use of computer systems provides a relatively large potential market to providers of electronic content or information. These providers may include, for example, advertisers and other information publishers such as newspaper and magazine publishers. A cost, however, is involved with providing electronic information to individual consumers. For example, hardware and maintenance costs are involved in establishing and maintaining information servers and networks. One source that can be secured to provide the monetary resources necessary to establish and maintain such an electronic information distribution network includes commercial advertisers. These advertisers provide electronic information to end users of the system by way of electronically delivered advertisements, in an attempt to sell products and services to the end users. The value of a group of end users, however, may be different for each of the respective advertisers, based on the product or services each advertiser is trying to sell and the class or classification of the user. Thus, it would be beneficial to provide a system, which allows individual advertisers to pay all, or part of the cost of such a network, based on the value each advertiser places on the end users the advertiser is given access to. In addition, advertisers often desire to target particular audiences for their advertisements. These targeted audiences are the audiences that an advertiser believes is most likely to be influenced by the advertisement or otherwise provide revenues or profits. By selectively targeting particular audiences the advertiser is able to expend his or her advertising resources in an efficient manner. Thus, it would be beneficial to provide a system that allows electronic advertisers to target specific audiences, and thus not require advertisers to provide an single advertisement to the entire population, the majority of which may have no interest whatsoever in the product or service being advertised or susceptibility to the advertisement.

U.S. Pat. No. 5,724,521, expressly incorporated herein by reference, provides a method and apparatus for providing electronic advertisements to end users in a consumer best-fit pricing manner, which includes an index database, a user profile database, and a consumer scale matching process. The index database provides storage space for the tides of electronic advertisements. The user profile database provides storage for a set of characteristics that corresponds to individual end users of the apparatus. The consumer scale matching process is coupled to the content database and the user profile database and compares the characteristics of the individual end users with a consumer scale associated with the electronic advertisement. The apparatus then charges a fee to the advertiser, based on the comparison by the matching process. In one embodiment, a consumer scale is generated for each of multiple electronic advertisements. These advertisements are then transferred to multiple yellow page servers, and the titles associated with the advertisements are subsequently transferred to multiple metering servers. At the metering servers, a determination is made as to where the characteristics of the end users served by each of the metering servers fall on the consumer scale. The higher the characteristics of the end users served by a particular metering server fall, the higher the fee charged to the advertiser.

Each client system is provided with an interface, such as a graphic user interface (GUI), that allows the end user to participate in the system. The GUI contains fields that receive or correspond to inputs entered by the end user. The fields may include the user's name and possibly a password. The GUI may also have hidden fields relating to "consumer variables." Consumer variables refer to demographic, psychographic and other profile information. Demographic information refers to the vital statistics of individuals, such as age, sex, income and marital status. Psychographic information refers to the lifestyle and behavioral characteristics of individuals, such as likes and dislikes, color preferences and personality traits that show consumer behavioral characteristics. Thus, the consumer variables, or user profile data, refer to information such as marital status, color preferences, favorite sizes and shapes, preferred learning modes, employer, job tide, mailing address, phone number, personal and business areas of interest, the willingness to participate in a survey, along with various lifestyle information. The end usher initially enters the requested data and the non-identifying information is transferred to the metering server. That is, the information associated with the end user is compiled and transferred to the metering server without any indication of the identity of the user (for example, the name and phone number are not included in the computation). The GUI also allows the user to receive inquiries, request information and consume information by viewing, storing, printing, etc. The client system may also be provided with tools to create content, advertisements, etc. in the same manner as a publisher/advertiser.

Use of Transactional Data for Marketing

In recent years, the field of data mining, or extracting useful information from bodies of accumulated raw data, has provided a fertile new frontier for database and software technologies. While numerous types of data may make use of data mining technology, a few particularly illuminating examples have been those of mining information, useful to retail merchants, from databases of customer sales transactions, and mining information from databases of commercial passenger airline travel. Customer purchasing patterns over time can provide invaluable marketing information for a wide variety of applications. For example, retailers can create more effective store displays, and can more effectively control inventory, than otherwise would be possible, if they know that, given a consumer's purchase of a first set of items, the same consumer can be expected, with some degree of probability, to purchase a particular second set of items along with the first set. In other words, it would be helpful from a marketing standpoint to know association rules between item-sets (different products) in a transaction (a customer shopping transaction). To illustrate, it would be helpful for a retailer of automotive parts and supplies to be aware of an association rule expressing the fact that 90% of the consumers who purchase automobile batteries and battery cables also purchase battery post brushes and battery post cleanser. (In the terminology of the data mining field, the latter are referred to as the "consequent.") It will be appreciated that advertisers, too, can benefit from a thorough knowledge of such consumer purchasing tendencies. Still further, catalogue companies can conduct more effective mass mailings if they know the tendencies of consumers to purchase particular sets of items with other sets of items.

It is possible to build large databases of consumer transactions. The ubiquitous bar-code reader can almost instantaneously read so-called basket data, i.e., when a particular item from a particular lot was purchased by a consumer, how many items the consumer purchased, and so on, for automatic electronic storage of the basket data. Further, when the purchase is made with, for example, a credit card, the identity of the purchaser can be almost instantaneously known, recorded, and stored along with the basket data. As alluded to above, however, building a transaction database is only part of the marketing challenge. Another important part is the mining of the database for useful information. Such database mining becomes increasingly problematic as the size of databases expands into the gigabyte, and indeed the terabyte, range. Much work, in the data mining field, has gone to the task of finding patterns of measurable levels of consistency or predictability, in the accumulated data. For instance, where the data documents retail customer purchase transactions, purchasing tendencies, and, hence, particular regimes of data mining can be classified many ways. One type of purchasing tendency has been called an "association rule." In a conventional data mining system, working on a database of supermarket customer purchase records, there might be an association rule that, to a given percent certainty, a customer buying a first product (say, Brie cheese) will also buy a second product (say, Chardonnay wine). It thus may generally be stated that a conventional association rule states a condition precedent (purchase of the first product) and a condition subsequent or "consequent" (purchase of the second product), and declares that, with, say 80% certainty, if the condition precedent is satisfied, the consequent will be satisfied, also. Methods for mining transaction databases to discover association rules have been disclosed in Agrawal et al., "Mining Association Rules Between Sets of Items in Large Databases", Proc. of the ACM SigMod Conf. on Management of Data, May 1993, pp. 207-216, and in Houtsma et al., "Set-Oriented Mining of Association Rules", IBM Research Report RJ 9567, October, 1993. See also, Agrawal et al., U.S. Pat. Nos. 5,615,341, 5,796,209, 5,724,573 and 5,812,997. However, association rules have been limited in scope, in the sense that the conditions precedent and subsequent fall within the same column or field of the database. In the above example, for instance, cheese and wine both fall within the category of supermarket items purchased.

U.S. Pat. No. 5,844,305, expressly incorporated herein by reference, relates to a system and method for extracting highly correlated elements (a "categorical cluster") from a body of data. It is generally understood that the data includes a plurality of records, the records contain elements from among a set of common fields, the elements have respective values, and some of the values are common to different ones of the records. In an initialization step, for each of the elements in the records, an associated value, having an initial value, is assigned. Then, a computation is performed, to update the associated values based on the associated values of other elements. The computation is preferably iteratively to produce the next set of updated values. After the computation is completed, or after all the desired iterations are completed, the final results, i.e., the updated associated values are used to derive a categorical cluster rule. The categorical cluster rule provides the owner of the data with advantageously useful information from the data.

Hierarchal Information Presentation

As the amount of information available to a computer user increases, the problem of coherently presenting the range of available information to the computer user in a manner which allows the user to comprehend the overall scope of the available information becomes more significant. Furthermore, coherent presentation of the relationship between a chosen data unit of the available information to the rest of the available information also becomes more significant with the increase of information available to the user. Most of the existing methods utilize lists (e.g., fundamentally formatted character-based output), not graphic models, to indicate the structure of the available information. The main problem associated with the use of lists is the difficulty of indicating the size and complexity of the database containing the available information. In addition, because the lists are presented in a two-dimensional format, the manner of indicating the relationship between various data units of the available information is restricted to the two-dimensional space. Furthermore, because presentation of the lists normally requires a significant part of the screen, the user is forced to reduce the amount of screen occupied by the list when textual and visual information contained in the database is sought to be viewed. When this occurs, the user's current "position" relative to other data units of the available information is lost. Subsequently, when the user desires to reposition to some other data unit (topic), the screen space occupied by the lists must be enlarged. The repeated sequence of adjusting the screen space occupied by the lists tends to distract the user, thereby reducing productivity.

One attempt to alleviate the above-described problem is illustrated by U.S. Pat. No. 5,021,976, expressly incorporated herein by reference, which discloses a system for enabling a user to interact with visual representations of information structures stored in a computer. In a system of this type, a set of mathematical relationships is provided in the computer to define a plurality of parameters which may be of interest to the user, which mathematical relationships are also capable of indicating a degree of correlation between the defined parameters and segments of information contained in a defined information system. In addition, an "automatic icon" with multiple visual features is provided to enable the user to visualize the degree of correlation between the parameters of interest to the user and the particular data unit stored in the computer that is being examined by computer. As the degree of correlation for a given parameter changes, the visual feature representing that parameter will change its appearance.

Another attempt to coherently present a large body of information to a computer user is illustrated by U.S. Pat. No. 5,297,253, expressly incorporated herein by reference, which discloses a computer-user-interface navigational system for examining data units stored in the memory of a computer system. In this navigational system, the user interface shows a continuous and automatically updated visual representations of the hierarchical structure of the information accessed. By using an input/output device to manipulate icons that appear in a navigational panel, the user can navigate through the information hierarchy. As the user traverses the information hierarchy, a node icon representing each level in the hierarchy accessed by the user is displayed. The user is also able to directly select any level in the information hierarchy between the entry point and the level at which the user is currently located.

Yet another approach to coherently presenting a large body of information to a computer user is "SEMNET," described in: Raymonde Guindon, ed., Cognitive Science and Its Applications for Human-Computer Interaction, (Hillsdale, N.J.: Lawrence Erlbaum Associates, Inc., 1988), 201-232. SEMNET is a three-dimensional graphical interface system that allows the users to examine specific data units of an information base while maintaining the global perspective of the entire information base. The SEMNET developers propose organizing component data units of an information base into various levels of hierarchy. At the lowest level of hierarchy, the most basic data units are organized into various sets, or cluster-objects of related information. At the next level of hierarchy, related cluster-objects from the lower hierarchical level are organized into a higher-level cluster-object Continuing in this manner, SEMNET achieves a hierarchical organization of the information base. In the graphic display, related data units within a cluster-object are connected by lines, or arcs. In addition, using a "fisheye view" graphic presentation, SEMNET displays the most basic data units near the chosen data unit but only cluster-objects of increasing hierarchy as the distance increases from the chosen data unit. In this manner, the user is able to visualize the organization of the information base relative to the chosen data unit. See, U.S. Pat. No. 5,963,965, expressly incorporated herein by reference.

U.S. Pat. No. 5,812,134, expressly incorporated herein by reference, relates to a system for interactive, dynamic, three-dimensional presentation of a database structure, seeking to allows the user to efficiently navigate through the database to examine the desired information. The system graphically depicts the organization of the information base as "molecules" consisting of structured parallel "threads" of connected nodes, each encompassing a specific aspect of the overall database. Within a given thread, the component nodes, which share a commonality of subject, are arranged in a natural, linear progression that reflects the organizational structure of the information subject represented by the thread, thereby providing the user with a visual guide suggesting the appropriate sequence of nodes to be viewed. By providing a hierarchical representation of the organizational structure of the entire database, the navigational system provides the user with both the "position" of the information unit being currently examined relative to the remainder of the database, as well as the information regarding the overall size and complexity of the database. The system also provides the user with the capability to define one or more "customized" navigation "paths" over the database, as well as copy and modify existing units of information. Thus, a taxonomy is constructed and employed to assist the user.

U.S. Pat. No, 5,774,357, expressly incorporated herein by reference, relates to a system that is adaptive to either or both of a user input and a data environment. Therefore, the user interface itself and/or the data presented through the user interface, such as a web browser, may vary in dependence on a user characteristic and the content of the data.

User Modeling

User modeling means to create a model of the user that contains information about the user that is relevant for a particular system. Thus, the user modeling system seeks to define sufficient characteristics of the user to determine the prospective actions or preferences of the user, and employ these characteristics to make predictions. Often, the user modeling system is used interactively with the user, facilitating the use of the system by intelligently predicting the user's inputs. On the other hand, a sufficiently accurate and constrained user model may also be used as part of an autonomous intelligent agent, i.e. a system that acts on behalf of the user to interact with other systems or persons.

The scope of the user model may include, for example, characteristics of the user which are independent of content, such as language, reading level, fields of expertise, physical impairments, and the like, as well as content specific characteristics, such as the user's taste and interests for motion picture entertainment, for example as part of a film recommending system, or the user's knowledge of a given academic subject, for an educational or testing system.

User preferences may be time dependent, and therefore diurnal or seasonal variations may be important factors in defining an accurate model of the user, i.e., the prediction of the intent and/or desires of the user in a respective context. Linear predictions, based on correlations, may be useful for extracting these patterns from observed sequences. More complex models, such as Markov models, may also be employed as appropriate. Often, the decision space is segmented into multiple operating regions, each defined so that the associated model is linear, e.g., MARS. Alternately, a nonlinear model, such as a neural network, may be implemented. Further, a combination of arbitrary type models and segmented decision space may be employed. A particular advantage of a segmented space is that the model for each respective segment is comparatively simpler, and may often be updated separately from other segments. The segmented architecture is especially advantageous where such models are implemented in applets, wherein the respective applet is simplified, and its execution speeded, by providing a narrower scope. Another advantage of a segmented user model architecture is that, when employed in conjunction with a collaborative filtering scheme, it may facilitate accounting for a greater range of user characteristics, while providing specific preferences.

It is also noted that at a plurality of user models may be employed, for example a content-independent and a content dependent type, with the outputs combined. These models may be segmented along common boundaries, or segmented independently.

Different systems use different techniques for constructing and implementing a user model. The simplest and most straightforward is a technique of a user survey, requiring some dedicated activity of the user toward defining the user model. A second technique monitors the activities of the user to detect patterns and actions indicative of user characteristics.

Intelligent agents can be constructed by monitoring or observing the user's actions with the system, and thereby determining characteristics, habits, tendencies or features of the user. Frequently visited pages, a request for an explanation of a technical term, often or seldom used links and functions are examples of things that can be examined. This is closely associated with machine learning, which means that the system learns the common actions and preferences of the user. Intelligent agents are often used for machine learning and this is a topic of artificial intelligence. This often requires the user to give the system some initial values about his or her knowledge, goals, experience etc. The initial values, or default values if not explicitly given by the user, are used for building a user model that will be altered when the agent discovers new things about the user.

U.S. Pat. No. 5,855,015, expressly incorporated herein by reference, proposes a system for retrieval of hyperlinked information resources which does not require a specific user query to locate information resources of interest, and which actively explores a hyperlinked network to present interesting resources to a user. Heuristics and relevance feedback may be used to refine an exploration technique, or to present resources of interest to a user. The proposed system continually adapts to changing user interests. A system for retrieval of hyperlinked information resources is provided which includes a user interface connected to a programmed microprocessor which is operative to explore the hyperlinked information resources using a first heuristic to select at least one information resource, to present the at least one information resource to the user via the user interface based on a second heuristic, to accept feedback from the user via the user interface, the feedback being indicative of relevance of the at least one information resource, and to modify the first and second heuristics based on the feedback. The patent also proposes a method for retrieval of hyperlinked information resources that includes exploring the hyperlinked information resources using a first heuristic to select at least one information resource, presenting the at least one information resource to the user via a user interface based on a second heuristic, accepting feedback from the user via the user interface indicative of relevance of the at least one information resource, and modifying the first and second heuristics based on the feedback. In one embodiment, the system utilizes a series of training examples, each having an associated ranking, to develop the first and second heuristics that may be the same, similar, or distinct. The heuristics utilize a metric indicative of the relevance of a particular resource to select and present the most relevant information to the user. The user provides feedback, such as a score or rating, for each information resource presented. This feedback is utilized to modify the heuristics so that subsequent exploration will be guided toward more desirable information resources.

The '015 system actively explores a hyperlinked network and presents a manageable amount (controllable by the user) of information to the user without a specific information query. Thus, the method allows selection of information of interest that may have been excluded by a precisely articulated query. Furthermore, rather than inundating the user with information selected from a general, broad query, the amount of information presented to the user is limited so as to minimize the time and effort required to review the information. This system provides ability to automatically learn the interests of the user based on a number of ranked training examples. Once exploration and presentation heuristics are developed, a hyperlinked network may be explored, retrieving and presenting information resources based upon the heuristics established by the training examples. The system is capable of continually adapting the exploration and presentation heuristics so as to accommodate changing user interests in addition to facilitating operation in a dynamic hyperlinked information environment.

U.S. Pat. No. 5,890,152, expressly incorporated herein by reference, relates to a Personal Feedback browser and Personal Profile database for obtaining media files from the Internet. A Personal Feedback browser selects media files based on user-specified information stored in the Personal Profile database. The Personal Profile database includes Profile Objects that represent the interests, attitude/aptitude, reading comprehension and tastes of a user. Profile Objects are bundles of key words/key phrases having assigned weight values. Profile Objects can be positioned a specified distance from a Self Object. The distance from the Profile Object to the Self Object represents the effect the Profile Object has in filtering and/or selecting media files for that user. The Personal Feedback browser includes a media evaluation software program for evaluating media files based on a personal profile database. The Personal Profile database is also adjusted based upon user selection and absorption of media files.

Another way of creating a user model is through the use of collaborative filtering. In this case, the user provides some initial information as well. For a collaborative filter, the user typically identifies himself or herself with a class of users, wither by predefined or adaptive categories. Thus, the emphasis of information gathering is not on the user's own knowledge or goals, but rather personal data such as age, profession or interests. The system then compares this user to other users and looks for users with similar answers to these questions. A user model is then created based on the profiles of similar users. Thus, collaborative filtering techniques typically require that a broad range of user characteristics be acquired and stored without aggregation, for later analysis and correlation to a given pattern.

According to one embodiment, during user interaction with the system, either including an explicit programming step such as a user survey, or through observation of the user, a user's characteristics are determined. Typically, it is too much of a burden on the user to explicitly obtain a complete profile. Therefore, any such profile is acquired in a goal-dependent or context sensitive manner. For example, a set of profiles are related by a decision tree. The user then explicitly or implicitly defines the necessary characteristics to traverse the decision tree to define an unambiguous profile, or to arrive at a set of compromises to define a hybrid profile. Since these profiles are goal-directed, the process of defining the profile is inherent in achieving the goal.

The particular profiles are, for example, defined by a logical analysis of the decision space, or defined by an analysis of a population of users, with each profile representing a cluster within the scope of the decision space. In the former case, it is often difficult to make presumptions about the user outside of the particular decision process; in the later case, by identifying a set of individuals within the population with broadly correlated characteristics with the user, it may be possible to infer user characteristics unrelated to the decision process.

Typically, after an explicit process of defining user characteristics, the system evolves into an adaptive mode of operation wherein the profiles are modified or updated to more accurately correspond to the specific user. Further, as the characteristics of the user become more fully available, collaborative filtering may be employed to make better presumptions regarding unknown characteristics of the user. It is also noted that the system preferably does not presume that the user has a consistent set of characteristics, and thus allows for changes over time and cyclic variations. Preferably, these changes or cyclic variations are analyzed and employed to extrapolate a future state.

A users' knowledge of the subject represented in the hypermedia is a particularly important user feature for adaptive hypermedia systems. Many adaptive presentation techniques rely on a model of the users' knowledge of the subject area as basis for adaptation. This means that an adaptive hypermedia system that relies on an estimate of the users' knowledge should update the user model when the user has presumably learned new things. Further, a preferred user model according to the present invention preferably also models decay of memory.

There are two common ways of representing users' knowledge in an adaptive hypermedia system. The most often used model is the overlay model that divides the hypermedia universe into different subject domains. For each subject domain in the hypermedia universe, the user's knowledge is specified in some way. The user's knowledge of a particular subject domain can be given the value known or unknown, or for instance a fuzzy semantic variable such as good, average or poor. On the other hand, a numeric or continuous metric may be provided. The user's knowledge may also be represented as a value of the probability that the user knows the subject. An overlay model of the user's knowledge can then be represented as a set of concept-value pairs, one pair for each subject. Overlay models were originally developed in the area of intelligent tutoring systems and student modeling, Greer, J. E., & McCalla, G. I. (Eds.): "Student Modeling: The Key to Individualized Knowledge-Based Instruction" NATO ASI Series F Vol. 125 (1993) Berlin: Springer-Verlag, but are also very useful for adaptive hypermedia systems. The main advantage of the overlay model is that users' knowledge on different topics can be measured independently. See, also Gaines, Brian R., and Shaw, Mildred L. G., "Concept Maps as Hypermedia Components", (Internet); Akoulchina, Irina, and Ganascia, Jean-Gabriel, "SATELIT-Agent: An Adaptive Interface Based on Learning Agents Interface Technology", In Anthony Jameson, Cecile Paris and Carlo Tasso (Eds), User Modeling: Proc. Of the Sixth Intl. Conf. UM97, Vienna, N.Y.: Springer Wein, N.Y. (1997); Benaki, Eftihia, Karkaletis, Vangelis A., Spyropoulos, Constantine D, "Integrating User Modeling Into Information Extraction: The UMIE Prototype", In Anthony Jameson, Cecile Paris and Carlo Tasso (Eds), User Modeling: Proc. Of the Sixth Intl. Conf. UM97, Vienna, N.Y.: Springer Wein, N.Y. (1997); Maglio, Paul P., and Barret, Rob, "How To Build Modeling Agents to Support Web Searchers" In Anthony Jameson, Cecile Paris and Carlo Tasso (Eds), User Modeling: Proc. Of the Sixth Intl. Conf. UM97, Vienna, N.Y.: Springer Wein, N.Y. (1997).

The other approach, apart from the overlay model, is the stereotype user model, in which every user is classified as one of a number of stereotypes concerning a particular subject or area. There can be several subareas or subjects, so one user can be classified as a different stereotype for different subjects. For instance, a novice stereotype, an intermediate stereotype and an expert stereotype can be defined for one subject in a system, and every user is therefore classified as one of an expert, novice or intermediate on that particular subject. This scheme is much simpler to implement but caries the disadvantage of not being able to tailor the appearance of the system to every individual user. Hohl, H., Bocker, H., Gunzenhauser R.: "Hypadapter: An adaptive hypertext system for exploratory learning and programming", User Modeling and user adapted Interaction 6, 2-3, (1996) 131-156, have shown that overlay modeling and stereotype modeling can be combined in a successful way. The stereotype model is used for new users to quickly create a reasonably good user model. Then the overlay model is used with initial values set by the stereotype model.

Users' goals often change from system usage session to session or even within a single session. The user's goal is often highly dependent on the kind of system employed. In educational hypermedia systems, the goal is often to learn a particular subject, or to solve a problem. In information retrieval systems, the goal can be to find a particular piece of information. In an institutional hypermedia system, the goal can be simply to do everyday work, which may be less easily described in generic terms. In systems where the set of goals is relatively small are unrelated to each other, Hook, K., Karlgren, J., Waern, A., Dahlback, N., Jansson, C. G., Karlgren, K. and Lemaire, B.: "A glassbox approach to adaptive hypermedia"; User Modeling and User-Adapted Interaction, 6, 2-3, (1996) 157-184, the system simply includes this goal in the user model. More complex and advanced systems, where goals cannot be separated distinctly, require more advanced inclusion and distinction of goals in the user model. One way of dealing with this is to create goal-value pairs for every possible goal in the user model, where the value is the probability that the user has this particular goal.

In some adaptive hypermedia systems, the user's background is considered relevant. The user's background means all information related to the user's previous experience, generally excluding the subject of the hypermedia system, although this exclusion is not necessary in all cases. This background includes the user's profession, experience of work in related areas and also the user's point of view and perspective.

The user's experience in the given hypermedia system means how familiar the user is with the appearance and structure of the hyperspace, and how easy the user can navigate in it. The user may have used the system before, but does not have deep knowledge of the subject. On the other hand, the user can know a lot about the subject, but have little experience of the hypermedia system. Therefore it is wise to distinguish between the user's knowledge and the user's experience, since optimal adaptations for each factor may differ.

The user's preferences are used in adaptive information retrieval systems mostly where they are the only stored data in the user model. Users' preferences are considered special among user modeling components, since they cannot be deducted by the system itself. The user has to inform the system directly, or by giving simple feedback to the system's actions. This suggests that users' preferences are more useful in adaptable systems than in adaptive systems. However, users' preferences can be used by adaptive hypermedia systems as well, as shown by Hook, K., Karlgren, J., Waern, A., Dahlback, N., Jansson, C. G., Karlgren, K. and Lemaire, B.: "A glassbox approach to adaptive hypermedia"; User Modeling and User-Adapted Interaction, 6, 2-3, (1996) 157-184. Hook et al. have found that adaptive hypermedia systems can generalize the user's preferences and apply them on new contexts. Preferences are often stored as numeric values in the user profile, contrary to the case for other data, which is often represented symbolically. This makes it possible to combine several users' preferences, in order to formulate group user models. Group models are useful when creating a starting model for a new user, where this user can define his or her preferences, and then a user model is created based on the user models of other users who are in the same "preference group".

Adaptive navigation support is used for helping the user to find the right paths through the hyperspace, by adapting the link presentation to the user's goals, knowledge, etc. Brusilovsky, P.: "Methods and techniques of adaptive hypermedia"; User Modeling and User-Adapted Interaction, 6, 2-3 (1996) 87-129, has found five different ways of adapting links to a user: direct guidance, sorting, hiding, annotation and map adaptation. Direct guidance means that the system suggests which links are best for the user to follow according to the user's goal, etc., in the user model. Sorting is an extension of direct guidance; all links are given a value according to how relevant they are for the user's goals etc. Hiding simply means that links that are considered not interesting for the user at the moment are hidden. In adaptive annotation systems, links are given a sort of comment about the current state of the node behind the link, for instance "not ready to be read yet". Map adaptation takes into account the human-computer interaction part. This is the only technique capable of fully adapting the layout of a page.

Machine learning and use of intelligent agents is a more useful technique than collaborative filtering, with respect to adapting the user interface to different users' needs. The reason for this is that the same user can have different needs at different times and therefore the system must respond to the user, and examine the user's actions, in order to understand what the user needs. In other systems that use user modeling, for instance in film recommending systems, the system already knows what the user wants and the interaction with the user is not as important.

U.S. Pat. No. 6,012,051 (Sammon, et al.), expressly incorporated herein by reference, relates to a system for processing user profiles to determine product choices likely to be of interest.

U.S. Pat. No. 6,006,218 (Breese, et al.), expressly incorporated herein by reference, relates to a method and apparatus for retrieving, sorting and/or processing information based on an estimate of the user's knowledge or familiarity with an object.

U.S. Pat. No. 6,012,052 (Altschuler, et al.), expressly incorporated herein by reference, relates to a method and apparatus for building resource transition probability models for use in various manners.

U.S. Pat. No. 6,014,638 (Burge, et al.), expressly incorporated herein by reference, relates to a system for customizing computer displays in accordance with user preferences. In accordance with the present invention, the user displays may thus be customized in accordance with a past history of use, including navigational choices, and personal characteristics and preferences.

U.S. Pat. No. 5,978,766 (Luciw), expressly incorporated herein by reference, relates to a system and method for suggesting nodes within a choice space to a user based on explicidy defined and/or observed preferences of the user.

U.S. Pat. No. 5,977,964 (Williams, et al.), expressly incorporated herein by reference, relates to a method and apparatus for automatically configuring a system based on a user's monitored system interaction.

U.S. Pat. No. 5,974,412 (Hazelhurst, et al.), expressly incorporated herein by reference, relates to an intelligent query system for automatically indexing information in a database and automatically categorizing users.

U.S. Pat. No. 5,970,486, (Yoshida, et al.), expressly incorporated herein by reference, relates to a method and apparatus for creating situation-dependent keywords, based on user characteristics and preferences, which are then used to define a query.

U.S. Pat. No. 5,963,645 (Kigawa, et al.), expressly incorporated herein by reference, relates to a system for receiving and employing personalized broadcast program metadata.

U.S. Pat. No. 5,801,747 (Bedard), expressly incorporated herein by reference, relates to a method and apparatus for monitoring a user's content consumption, to infer user preferences therefrom.

U.S. Pat. No. 5,758,259 (Lawler), expressly incorporated herein by reference, also relates to a user preference profile determination system which monitors user activity.

U.S. Pat. No. 5,945,988 (Williams, et al.), expressly incorporated herein by reference, further relates to a similar system for dynamically updating inferred user preferences based on user activity.

U.S. Pat. No. 6,005,597 (Barrett, et al.), expressly incorporated herein by reference, relates to a system and method for monitoring user content consumption and creating a dynamic profile based thereon, which is then used to sort future available content.

U.S. Pat. No. 5,973,683 (Cragun, et al.), expressly incorporated herein by reference, relates to a system for the dynamic regulation of television viewing content based on viewer profile and viewer history.

U.S. Pat. No. 5,946,490 (Lieberherr, et al.), expressly incorporated herein by reference, relates to an automata-theroretic approach compiler for adaptive software. Such a compiler could be used, for example, to produce customized applets for users representing a set of search results, or incorporating user profile data.

See, also:

Boyle C. and Encarnacion A. O.: "MetaDoc: an adaptive hypertext reading system"; User modeling and User-Adapted Interaction, 4 (1994) 1-21.

Brusilovsky, P., Eklund, J.: "A study of user model based link annotation in educational hypernedia"; Journal of Universal Computer Science, Vol. 4 No 4 (1998) 429-448.

Chin, D.: "User Modeling in UC: the Unix Consultant"; Proceedings of the CHI-86 Conference, Boston (1986)

Moore, J. D. & Swartout, W. R.: "Pointing: A way toward explanation dialogue"; Eight National Conference on Artificial Intelligence, (1989) 457-464.

[AVANTI homepage ] zeus.gmd.de/projects/avanti.html

Fink, J., Kobsa, A., Schreck, J.: "Personalized hypermedia information provision through adaptive and adaptable features: User modeling, privacy and security issues" zeus.gmd.de/UM97/Fink/Fink.html

Eftihia Benaki, Vangelis A. Karkaletsis, Constantine D. Spyropoulos, "Adaptive Systems and User Modeling on the World Wide Web", Proceedings of the workshop, Sixth International Conference on User Modeling, Chia Laguna, Sardinia, 2-5 Jun. 1997

Brajnik, G., Guida, G., Tasso, C., (1990): User modeling in Expert Man-Machine Interfaces: A case study in Intelligent Information Retrieval, in IEEE Transactions on systems, man, and cybernetics, 20:166-185.

Brajnik Giorgio and Carlo Tasso, (1994): A shell for developing non-monotonic user modeling systems in International Journal of Human Computer Studies, 40:31-62.

Croft, B. and Thompson, R., (1986): An overview of the IR Document Retrieval System, in Proceedings of the 2nd Conference on Computer Interfaces and Interaction for Information Retrieval.

Karkaletsis, E., Benaki, E., Spyropoulos, C., Collier, R., (1996): D-1.3.1: Defining User Profiles and Domain Knowledge Format, ECRAN.

Kay, J., (1995): The urn toolkit for Cooperative User Modeling, in User Modeling and User-Adapted Interaction, 4:146-196.

Jon Orwant, (1993): Doppelganger Goes to School: Machine Learning for User Modeling, M.Sc. thesis at MIT.

J. Orwant, "For want of a bit the user was lost: Cheap user modeling", MIT Media Lab, Vol. 35, No. 3&4 (1996).

Rich, E., (1983): Users are individuals: individualising user models in International Journal of Man-Machine Studies, 18:199-214

Collaborative Filters

Collaborative filtering is a process that seeks to determine common habits for a group of individuals having a common characteristic. The information is commercially valuable, for example, because knowing what a consumer has purchased, it is possible to statistically predict what else he or she might buy, and to suggest such a purchase to the consumer. This collaborative filtering is preferably used in conjunction with a past history of the user himself, providing a powerful tool for predicting consumer behavior and preferences.

Collaborative filters presume characteristics of the user based on a class identification of the user. A collaborative filter may be adaptive, meaning that it is updated based on actions subsequent to the classification of the user relating to the success or quality of the classification. According to an adaptive embodiment of a collaborative filter of the present invention, therefore, it is preferred that the client system, either concurrently with use of the system by the user, or subsequently, transmit to the server sufficient information to update the collaborative filter for more accurately classifying the user and/or for more accurately defining the characteristics of a respective classification.

Collaborative filtering is often used in systems that continuously present new things to their users such as film recommending systems and personally adapted news. If the user does not like one suggestion from the system, he or she tells the system that this material is not interesting and the user profile is updated. Typically, the weight afforded to adaptive features is low, and indeed the feedback from one user is just one input to the system, and thus will typically not cause a major change in the user profile, since most of it still is valid. Systems that adapt the user interface to different users' needs often need to give the user more control over the adaptation process. It is much more difficult to predict the user's preferences correctly in these systems since they may vary with time. For instance, the user's knowledge of a subject can be a component in the user model, and it is hard for the system to predict exactly when a user has learned something new. The system needs some help from the user, and what the user says is more important than the current user model. The user model has to be modified completely to what the user has said. Therefore, intelligent agents and machine learning are preferred in these systems.

Collaborative filters and user profiles necessarily require that personal user information be employed. This personal information may include private user information, such as demographics, preferences, past purchase history, media consumption habits, and the like, or confidential information including trade secrets, or information otherwise not intended for publication. The unrestricted release and distribution of this private user information, or the risk of dissemination, is typically undesirable, from the user's viewpoint. In the case of collaborative filtering systems, this information must be stored centrally, thereby creating a risk of breach. In the case of adaptive personal profile systems, client-side filtering may be employed; however, this necessarily entails transmission of a greater amount of information to the user than is presented to the user. Client-side filtering requires that all information be transferred to the client system, which is often expensive or untenable. In general, any time valuable personal profile information exists, even in when physically in a client system, a risk of misuse or misappropriation exists.

The release and distribution of private user information, such as demographics, preferences, past purchase history, media consumption habits, and the like, typically is avoided, and may be limited by law or agreement. Therefore, one option available for filtering or processing information based on this information is at the client system, where the private information need not be released or distributed. For example, see U.S. Pat. No. 5,920,477, expressly incorporated herein by reference, and Metabyte Inc., www.mbtv.com, which disclose systems for determining user preference profiles for television programs, implementing a client-side filter. However, this requires that all (unfiltered) information be transferred to the client system, for subsequent filtering, which is often expensive or untenable. Further, this requires computational resources at the client for filtering the content. However, in various circumstances, such techniques may be employed.

SUMMARY AND OBJECTS OF THE INVENTION

The present invention provides a personal services infrastructure.TM., which unifies the visual environment through the use of stylized taxonomic trees and timelines ("maps"). The mapping process transcends the cumbersome "file folder" technology offered for many years by the leading operating systems and browsers, with gains in understanding, flexibility and compactness. A particular aspect of the present invention is that with an intelligent organization of information, supplemental information, i.e., information not originally part of the data being organized or displaced from its proper location within a classification system, may be presented with properly organized information.

The present invention provides, according to a preferred embodiment, a method of visualization of a set of elements in a computer graphic interface, comprising defining a hierarchy of objects, each hierarchal level within the hierarchy having at least one object, the at least one element having one parent hierarchal object and optionally a set of child objects, with a set of content objects populating the hierarchy; defining, based on a user input, a selected object within the hierarchy for examination; and generating a display on the graphic user interface, presenting the selected object element and any child objects thereof; a representation of parental objects within the hierarchy, with a representation of a hierarchal representation thereof; wherein each of the parent and child objects is associated with a hyperlink, a selection of a respective hyperlink serving to transform that element into the selected element, wherein when a content object is selected, an associated set of related objects is displayed. Therefore, the present invention provides a graphic hyperlink hierarchy providing, with a display of a content object, a display of an associated object.

Preferably, the associated object is defined by a process of collaborative filtering. For example, the associated object may be an advertisement, offer of a product or service for sale, or a set of information. Preferably, an economic motivation is present for defining the associated object, for example, a sponsor or other party might seek, based on an identification or special characteristics of the user or the class of content, to communicate with the user.

The associated object is preferably contextually appropriate, although in some instances, the associated object will have no apparent relation. Thus, for example, political ads may appear in blanket campaigns, regardless of the context, except that such ads may be directed more toward adults than children.

The present invention also provides a method for a user, having found a Web or non-Web resource of interest and relevance, to find similar resources. According to this method, the user selects the preferred resource. That resource is then mined for concepts and phrases, of which the most prominent are presented to user as a second step. User selects which concepts are relevant to finding similar resources. A search or metasearch is then carried out, in which these particular concepts are searched for.

A parallel process is provided for the user to find a similar product to a preferred product. A special feature of this process is that characteristics of the product or service are mined for, including for example nature of product, price, quality, warranty features, and service. The user is then asked to rank or rate the importance of those features that are important to user. A search of metasearch is then carried out. The user is presented with a selection of similar products or services. Alternatively, or in combination with the above method, user's query may be "broadcast" over a computer network or otherwise and invitations made to potential sellers to make an offer of the same or similar product or service, with each seller competing to make the best offer. User would then be offered a selection from these competing offers. As a special feature of this competition, offers could be displayed as made, with the possibility of a fixed offering period or an open offering period. Generally speaking, a user having focused on a particular item that user is considering for purchase, the broadcast of this information may lead to a user receiving a better offer, in terms of user's preferred ensemble of product characteristics, or elicit product alternatives that user may not have known of or could not have accessed. The items offered may be displayed in the portion of the visualization where collaborative filtering content of commercial suggestions are usually displayed.

The object is preferably defined by a query, for example a Boolean search of the content represented in hierarchy of objects.

This supplemental information is provided either to enhance the user's experience or results, or to provide revenues. Exemplary revenue producing transactions include advertising and electronic commerce opportunities.

The present invention is preferably implemented using a web browser, such as Netscape Navigator or Microsoft Internet Explorer, using hypertext markup language (HTML) and/or extensible markup language (XML), and optionally helper applications, JAVA applets, Visual Basic applets or programs (e.g., OCX), or other known program constructs. The browser typically resides on a client system, having a user interface, processor, storage, and a connection to a communications network. The database server is typically remote from the user, and services a large number of users. See, "The Java.TM. Language Environment: A White Paper", James Gosling & Henry McGilton, www.quant.ecol.klte.hu/javatjava_whitepaper.sub.-- 1.html (et seq.). The client system is typically capable of storing and processing information locally, while the communications network connection may prove rate limiting. Therefore, it is preferred to employ the storage and processing capacity of the client system to reduce the information that must be transmitted. Further, the browser typically provides a document page model for information display, which may be quite limiting. The present invention therefore preferably provides an application or applet for providing advanced display and interaction facilities for interfacing the user with the information from the server.

An applet is a program designed to be executed from within another application, for example a JAVA applet executing within the JAVA Virtual Machine (JVM). Unlike an application, applets typically cannot be executed directly from the operating system; in other words, the applet typically relies on resources that are not native to the operating system, but rather are supplied by the host application. When OLE (object linking and embedding) techniques are employed, an appropriately designed applet can be invoked from many different applications. According to a preferred embodiment of the invention, enhanced functionality is provided by a downloadable applet that does not require a user-install process or lengthy download times.

Zoomable Nested Nodal Hierarchies

The system according to the present invention preferably provides an improved user interface which may include the visual presentation of information in a form that is (a) hierarchal, that is, organized in levels of generality according to a scheme, (b) nested, that is set together in groups depending upon associated characteristics, (c) zoomable, in the sense that a user, in varying by at least one degree the level of generality, also varies the view (not necessarily continuously zoomable, like a lens, but sometimes stepwise zoomable), and/or (d) nodal, in the sense that points are presented to user as hyperlinks to a particular level of generality. Such representations including all these characteristics are called zoomable nested "nodal networks". A nodal network consists of a set of "nodes", or discrete and defined objects, connected by links, each link typically having two ends and defining relationship between the linked objects. The term zoomable infers that the nodal network may be examined and convey useful information on a plurality of different scales, and thus may be represented to the user at such different scales. These elements define a hyperlink tree, i.e., a nodal network wherein each node represents and identifies an object, the object being generally accessible by selecting a respective node, and wherein the zoom provides a selective disclosure of underlying nodes based on a degree of scrutiny or "zoom". Such zoomable nested nodal networks resemble a traditional botanical taxonomic tree, and thus these networks may be called "trees". According to the present invention, however, the rules and tenets of taxonomy are not absolute, allowing a greater degree of flexibility for display, representation and manipulation of the objects and information represented. Of course, a formal taxonomy may be adopted.

One hyperlink tree, a Hyperbolic Tree.TM. (Inxight Software Inc., Palo Alto Calif.), developed at Xerox PARC, is disclosed in John Lamping, Ramana Rao, and Peter Pirolli, "A Focus+Context Technique Based on Hyperbolic Geometry for Visualizing Large Hierarchies", CHI 95, _.acm.org/sigchi/chi95/proceedings/papers/jl_bdy.htm. See also, www.inxight.com,_.inxight.com/News/Research_Papers_Files/Z-GUI_Article. pdf?.

An alternate hierarchal representation of information is provided by TheBrain.com, Santa Monica, Calif. 90404, www.thebrain.com, which has developed a dynamic information presentation applet showing hierarchal links between data elements, which may include hyperlinks to associated resources. More recently, TheBrain.com has developed an Open Directory Project search service for presenting search results within their applet framework. This is not believed to be prior art to the present invention.

In additional to multiresolution representation and analysis (e.g., zoomable viewing), there are other options that may be predefined or defined by the user with respect to information or organizational display. For example, the tree structures may be represented as horizontally or vertically oriented taxonomic trees, a radially oriented tree, an outline with indentations, a conceptual map, a 3D conceptual map, with a virtual third dimension, such as "height", added to the image, an n-space map, with multiple degrees of freedom represented in various visual or other sensory means, or the like.

According to the present invention, a set of information may be transmitted from the server to the client, for presentation to the user. The information may be classified according to the ultimate taxonomy, regardless of the level of analysis employed by the user, or may be classified only to a lesser level of granularity, for example a level specified by the user or adaptively determined based on the user query, user profile, and the information content retrieved.

U.S. Pat. No. 6,014,671 (Castelli, et al.), expressly incorporated herein by reference, relates to an interactive retrieval and caching of multidimensional data using view elements. According to this patent, view elements include node elements and transition elements between nodes.

Protection of Personal Profile Information

The present invention proposes a number of means for minimizing the risk of release of personal information. For example, the present invention provides a set of Intelligent Agents, wherein the private information forming the basis for agent action is encrypted using a secure encryption method, either embedded or associated with the Intelligent Agent, or securely transmitted to it. The encryption technique may be of any suitable known type, for example public key/private key techniques, RSA algorithms, elliptic key techniques, etc. The Intelligent Agent preferably is provided as an applet, either integral with the user interface applet or associated with it. On the other hand, an Intelligent Agent applet may also physically reside at a server location, being shielded from interrogation or analysis by a combination of so-called firewall protection, encryption, and logical restrictions on the quantity and nature of information released. The Intelligent Agent further is preferably protected from being probed to methodically determine the included private information, such as by generating spurious responses or "pseudorandom noise" (an apparently random yet predictable pattern based on a complex algorithm), which may be filtered at the client system, and by storing and analyzing a history of usage to detect and thwart hacking. While there may be cryptographic methods for breaching these types of security measures, such methods are computationally intensive and therefore may be more difficult than other surreptitious methods of obtaining the private information.

The present invention thus provides an intelligent agent system, wherein the user private information is encrypted using a secure encryption method, either embedded in a custom intelligent agent for each user, interactively and securely transmitted to it. The encryption scheme may be of any suitable known type. In this case, the server stores a set of user-specific intelligent agent applets or data files, which are called upon as required to provide or supplement information about the user. These intelligent agents or data files may be adaptively updated, based on recent feedback from the user or respective use, with the updated agent or data file encrypted and the raw data purged. Therefore, the intelligent agent applet may physically reside at a server location, while being shielded from interrogation or analysis by either a secure firewall, encryption, or both. The intelligent agent further is preferably protected from being probed to determine the private information, such as by generating spurious responses (or being formulated with a portion of spurious data) or producing "pseudorandom noise" (an apparently random yet predictable pattern based on a complex algorithm), which may be filtered at the client system based on a knowledge of the complex algorithm, and by storing and transmitting a history of usage to detect tampering. While there may be cryptographic methods for breaching this security, such methods are computationally intensive and therefore may be more difficult than other surreptitious methods of obtaining the private information.

Hierarchal/Taxonomic Organizational Schema

According to the present invention, an information retrieval hypermedia system is provided which includes an adaptive user interface, in which presented search results contain hierarchal associations of sets of documents, wherein respective hierarchal associations are based on user-specific data distinct from the formal query itself. Thus, for example, according to a user hierarchal schema, documents providing similar or related information are classified together, wherein this similarity or relatedness is not defined intrinsically in the query. Further, the hierarchal schemas may be persistent, and applied to results of multiple distinct queries. Alternately, a user hierarchal schema may be specifically defined for a particular query or topic of inquiry. By providing an organization of query responses, users may define a broad query scope that encompasses a desired topic, but may also encompass other topics, either intentionally or because the user is unable to a priori precisely define the query scope. Often, a query produces a large number of hits, and the user has difficulties finding relevant information in an unorganized set of query results. By presenting a linkage between similar documents, the user may not have to go through all responses to the query (search hits), but can skip many documents after having characterized the group or the contents of the group, e.g., read one of the documents.

An aspect of the invention therefore provides means for the user to refine the search criteria in order to improve the precision of search results returned. Preferably, this is an interactive process in which packets of information are communicated between the client and server, although it is possible to conduct this process solely on the client system. Where an interactive scheme is employed, it is possible to transmit, for example, marketing information to the user (e.g., banner ads) with each downloaded packet, or otherwise communicate information in spare or otherwise available bandwidth during this process. These added opportunities may be used, for example, to subsidize the use of the system that allows the user to define or refine the query.

This intelligent assistance preferably involves an interactive communication between the user and search engine, wherein a context, e.g., semantic taxonomic placement, of the search query is successively defined and refined. Preferably, after the context of the query is defined, the user is presented with a hierarchal tree of contents, i.e., a branched hierarchical graphic representation of the information and linkages, for confirmation. In the event that the relayed context is accurate, a simple confirmation is accepted. On the other hand, where the context is not accurate or of inappropriate scope, the user may change or refine the context In this way, the number of complete database searches is reduced, and the results tailored to the user's expressed requirements.

By providing a hierarchical tree of contexts, the user is prompted to select or accept the narrowest definition scope of the query. In most instances, this will result in a narrower search than a simple one or two word query, but it may also provide an intelligent means for broadening the scope while avoiding an undue number of returned irrelevant hits.

The hierarchical tree of contexts may be presented to the user in a bounded rectangular box, for example showing three levels of hierarchy, with a single node at the highest level, a set of intermediate level nodes defining a range within the generic (highest level) taxonomy, but not necessarily being exhaustive for that level. A selected set of lower level nodes are also presented, which also need not be exhaustive, and in fact, it is preferred that this level be truncated if necessary in order to reduce visual clutter. The nodes are preferably connected with line segments. In the event that an ambiguity is presented, or otherwise the user is to be presented with multiple discontinuous representations of the taxonomy, each may be presented in a separate bounded rectangle. It is understood, of course, that the tree structure need not be presented in a rectangle, and indeed alternatives to a visual tree are also possible.

Preferably, each node within the structure is active, so that a graphic manipulation of the node in a web browser may be detected. This node need not represent a hyperlinked URI, however, and means are preferably provided for selection of one or more nodes by the user without intermediate screen information refresh. In fact, in some embodiments, only terminal child nodes of a hierarchal object are hyperlinked, for example to URIs or a search results page object, with all higher order nodes being locally interpreted at the client system.

In one embodiment, the nodal representation is presented as a graphic map, wherein a Cartesian coordinate of a pointing device is transmitted to the server to indicate a manipulation thereof. The server correlates the coordinates of manipulation with the graphic element at that point Alternately, an applet may be provided to generate the nodal graphic. In this case, the helper application may intercept and process manipulation of the graphic, without requiring intervention of the server. Further, the applet may locally store a larger portion of the taxonomic structure than is displayed, which will allow faster refresh and improved real time interactivity, at the potential expense of a longer initial activation latency. Preferably, an applet locally stores a set of higher levels of the taxonomic hierarchy, as well as a cache of recently used lower levels. As the taxonomy requires updating, the applet may communicate with the server. In addition, the applet may provide further graphic information to the user, for example relevant ads or navigational hints, without interrupting the user's interaction with the nodal representation.

Typically, the taxonomic contexts will be semantic, e.g., a verbal expression of an idea. The hierarchical taxonomy will therefore represent, at least in part, a linguistic analysis of the proposed query. Typically, the taxonomy will include a single linguistic concept, which will be distinguished from other concepts, even those with an identical literal expression. In some cases, a multi-term query will represent an attempt to define a single linguistic concept. In that case, a single taxonomic classification will be defined, and the search formulated to retrieve records corresponding to that concept. Often, it is not or will not be possible to determine a context of a record a priori, i.e., during the indexing process. In that case, the record may be analyzed as a part of the search process to determine if it meets the search criteria or otherwise the ranking it should receive for relevance. On the other hand, it may be impossible to determine automatically (or manually) the context of a record. In that case, a set of rules may be applied to deal with this case. For example, the user may determine that these records should be retrieved, should be ignored, should be given a high or low ranking, or otherwise. The rules may also take into consideration the quantity and nature of other records retrieved (or excluded). Typically, it is desired to maintain a stateless condition, therefore, once the search query is executed, the results should be downloaded to the client, or explicitly defined in a URI. The database server, therefore, typically does not retain the query response for an extended period. See, e.g., U.S. Pat. No. 6,012,053 (Pant, et al.), expressly incorporated herein by reference. However, the server may retain search results for a period of time, for example 5 days, to allow the user access to prior search results from the server (e.g., allowing the user to employ multiple client computers or diskless computers), and to provide data for the server to establish user profiles.

In another aspect of the invention, the search query need not be limited to linguistic concepts. Therefore, the search may involve images, video, audio, or other types of data. In this case, the taxonomy may be based, for example, on characteristic patterns or attributes of the data sought. It is noted that there are a number of systems available that support non-verbal data access and retrieval. These include the QBIC system from IBM, products from Virage, Informix, Excalibur, Magnifi, Muscle Fish LLC, and a number of other entities. In fact, the interactive search definition according to the present invention is advantageous where a simple verbal search query is untenable, such as in searching non-linguistic data.

In some instances, the context of the query will not be fully or appropriately defined by a predetermined linguistic taxonomy. Therefore, the system may define a temporary or artificial taxonomy. This taxonomy may be based, for example, on an analysis of the records (or a select subset) themselves, or by the user during the interactive process. It is also possible for a user to store a preference profile, which may include, for example, taxonomic or heuristic concepts. The database server, therefore, may reference this profile in responding to the query. In known manner, this profile may be stored locally on a client system, e.g., as a cookie, or remotely, in a server in a file referenced to the user. For example, U.S. Pat. No. 5,895,471, expressly incorporated herein by reference, relates to a system for use with mobile, storage constrained clients, which stores hypermedia links such as Uniform Resource Locators (URL), used to identify and control access to resources on the network, on a server remote from the client device. Another system provides a globally unique identifier (GUI) to track users across secure and insecure networks. See, U.S. Pat. No. 5,966,705, expressly incorporated herein by reference.

A user may, for example, be provided with a personal web page, including a variety of information. In some instances, this information will be personal, and will therefore be maintained in secrecy, for example requiring passwords and/or support for encryption (e.g., secure socket layer [SSL] communications). Advantageously, the information associated with this web page may be updated and enhanced automatically, to represent a history of use by the user. Because this web page is maintained separately from the database server, it is accessible to a variety of servers, and further may be referenced by URL. Therefore, this scheme allows an on-line "memory" and persistence of complex parameters even where the system is otherwise stateless. This scheme differs from the simple use of Internet cookies, in that the file may be stored remotely, and is therefore not encumbered by the communication link between the user and server. Further, it is possible to perform analysis, e.g., stochastic analysis, of the profiles of a number of users, in order to improve the performance of the system. These files are "personal", which mean they are linked to the identity of the particular user, rather than the particular machine from which he communicates.

In the taxonomic representation, which, as discussed above, is preferably a tree structure, each node may be a hyperlink, meaning that a selection of that node indicates a reference to another data object or URI. See, e.g., U.S. Pat. No. 6,018,748 (Smith), expressly incorporated herein by reference. In some instances, the selection of a node will be employed to define a refinement of the taxonomic definition. In other instances, the selection of a node may point directly to a data element. Thus, for example, where the taxonomic definition is sufficiently specific, the selection of a node automatically calls a URI, which may initiate a search in a search engine or call a specific web page. On the other hand, the user may select a group of nodes to define a concept cluster. Graphically, the user may circumscribe a set of nodes, potentially across multiple taxonomic levels or even discontinuous through the taxonomy, to define the context. Where a node or group of nodes represent a search definition, the definition may be directly derived from the taxonomy, or it may represent the labors of human experts who translate the context of a node into an optimized query. Likewise, the query string itself may vary depending on the search engine referenced. Further, the search string may also vary in dependence on a "sophistication" or "role" of the user.

The user may create de novo, modify or extend a predefined taxonomy based on use or particular requirements. Therefore, the present invention provides a generic taxonomic structure for the organization of knowledge, and in particular computer and Internet platformed information, and to which a set of new, predefined or extensible definitions may be associated. From a commercial point of view, each person's activities and interests could be seen as hot spots on a predefined taxonomic map. To use a visual metaphor, a set of transparencies, each representing a taxonomic map of a person's interests and activities overlaid upon one another, would show darkened areas similar to population clusters in a population map. This metaphor could also be translated into a statistical model of groups of people sharing common interests for the purpose of sales and e-commerce. The present invention therefore encompasses the collation and use of such taxonomic maps of the activities and interests of specific populations. These may be used, for example, to generate custom sales catalogs, either printed or on-line.

The present invention also encompasses selections of information, e.g., customized catalogs, generated for individuals or population groups, based on the structure or statistical density of populated nodes on a taxonomic classification of interests and activities.

The present invention also encompasses the idea that there are certain domains of knowledge where a user may never have an interest. Thus, a child may have little interest in real estate listings. It provides the opportunity for user to exclude certain categories of information on a durable basis. This may be accomplished through a keyword methodology--certain keywords are related to real estate listings and suggest non-relevance, or by taxonomic exclusion, so that certain branches of a taxonomic tree are durably or semi-durably excluded. This approach to date has been limited to exclusion of "adult" material, but it has a much broader utility. In distinction to "smut filters", the system according to the present invention is therefore adaptive, providing individualized filters and inclusion/exclusion (or ranking) criteria. In fact, these criteria may be context sensitive, such that application of a criterion is dependent on the history of use (recent and/or long term), data environment of the system, e.g., explicitly and/or implicitly entered information and automated responses thereto, and/or status of the system, e.g., responsive to the tasks presently executing on the system. Therefore, according to the present invention, filters need not be absolute, such that in the aforementioned example, an otherwise relevant response to a query need not be excluded simply because it contains words which are likely indicative of a real estate listing, if the response is otherwise material.

This, of course, raises the issues of competing and cooperative filters. According to the present invention, an intelligent decision may be made dependent on outputs of a plurality of semantic, taxonomic, or other types of filters. Of course, a rigid filter rule, such as a "smut filter", may also be established within the same framework.

In some instances, a predetermined taxonomy is insufficient to finely granularize the set of results returned. In addition, the user may not be able a priori to classify the results without first examining them. In these sorts of instances, it may be desired to automatically classify documents into subsets of records of reasonable number. Thus, a relatively large set of objects responsive to a query may be automatically analyzed to determine common characteristics and categorized into mutually exclusive (or reasonably so) subsets thereof. The system may then define these distinctions as temporary (or permanent) taxonomic classifications. The user may then review these derived classifications, generated based on the content of the objects, or the objects therein. The automated classifications may also provide extrinsic distinctions, e.g., commercially motivated distinctions, rather than purely intrinsic content related distinctions.

As discussed above, the recently relevant portions of the taxonomy may be cached by an applet or helper application, and therefore these cached portions may include the derived taxonomy. Typically, the artificial taxonomy may be difficult to automatically integrate into a predetermined taxonomy. In that case, the system may offer the user the opportunity to manually define a taxonomic relation of a new or artificial taxonomic classification. Preferably, this opportunity is presented asynchronously with other requests of the server. Further, the applet may organize and defer such tasks. In fact, the applet may assist the user in organizing information extrinsic to the searching system, so that the user's available information base, from multiple sources, is coherently organized.

In a preferred embodiment, a taxonomy may be defined based on a commercial or industrial interest. Thus, the system according to the present invention may be applied to catalogs and specialized databases. Further, the taxonomy may be defined as a set of nodes, each node representing a different resource. For example, in an electronic (on-line) commerce system, each vendor may be represented as a node within the taxonomy, based on the products or services offered, client profile, and other factors, such as a priority rating. Therefore, the user may be presented with a plurality of potential taxonomic systems, depending on an initial interaction with the system. The user may initially indicate that he or she seeks to purchase an item. Therefore, the taxonomy selected will relate to goods for sale. In that case, a particular item offered within the system may ultimately have a plurality of taxonomic classifications, depending on a path the user takes. In other systems, the taxonomy is constrained such that no item may be classified more than once. An example would be linguistic searches, wherein a single "meaning" for a search term is desired, and the taxonomy defines the meaning.

Certain parts of the taxonomy might be made available on a commercial basis. For example, under automobiles, American automobiles, there might be a portion of the tree with Chrysler, under which might be sports utility vehicles (SUV), trucks, cars, and under sports utility vehicles might be Durango and Cherokee, where this inset in the taxonomy is paid for by the manufacturer or distributor on a fixed fee for view or click-through basis, or a combination of these approaches. The advertiser would know that a person entering its portion of the taxonomical tree really wants to see this particular product or aspect of its business, which should command a premium fee or click-through charge. To maintain integrity with the users, the paid portions of the taxonomy could be differentiated with a distinguishing typeface or color, could be outlined as commercial, or otherwise differentiated, to separate commercial and non-commercial portions of the taxonomy. In other cases, the probability of a user responding in the desired way to the ad is not exceedingly high. In that case, the cost per impression could be lower, or a higher valued ad substituted. The advertising rates may therefore be variable, and even computed according to a continuous formula, based on the characteristics of the user, the present search and past history of the user, and possibly other factors, for each imprint or user.

In the same manner, premium content, i.e., information objects that are available only by subscription or through payment, may also be highlighted and/or segregated from free or basic content.

The taxonomy may also include a hybrid representation, especially where commercial subsidies are a factor. Thus, where a user is "shopping", the interactive search process is purely of a commercial nature, and is optimized accordingly. This optimization may be such as to maximize revenues for the search engine proprietor, or maximize sales profit for the vendor. On the other hand, where the user seeks "content", rather than to purchase, the process may be subsidized by seeding the visual displays presented to the user with advertisements. Preferably, these advertisements are targeted to the user, for example based on the search premise, an imputed user profile or set of characteristics, or an identification of the user. Thus, for example, relevant commercial elements may be interposed in the taxonomnic structure of the content. Alternately, banner ads may be provided, associated with the content displayed, the user, or otherwise to the process in which the user is involved.

This commercially subsidized portion of the information may be subjected to various filters, limits and compensation attributes. Thus, a user may wish to avoid all extraneous information, at the cost of usage fees, subscription payments, or other model for compensation of the service provider. Likewise, the amount of sponsor information may be limited, either at the server transmitting the information, or at the client system.

Revenue Models

A further aspect of the invention relates to revenue models, which may be defined, based on the advanced functionality of each respective system. For example, in the process of defining user characteristics for the system, the user conveys valuable information about himself. This information may be used, for example, to define, on a general level, a set of products and services in which the user may be interested. This information may be used internally within the system, or sold to marketing concerns, as permitted by the user, usage agreement, regulation and law. Advantageously, the system "tests" hypotheses by requesting feedback from the user relating to generalizations and specifications that are made. Thus, the function of defining the characteristic of the user may be enhanced through cooperation of the user. The benefit to the user of allowing these characteristics to be ascertained will be tangible and immediate, so the user will likely not object. The value of this data, in turn, may be returned to the user, in whole or part, by monetary remuneration, subsidy for search activities and/or reduced 'search costs" for items of interest to the user.

Since a significant cost in Internet searching involved retrieval of query responses, by narrowing a scope of a search, it is possible to reduce the costs by limiting the information which must be delivered. Further, the value of the identification of a user characteristic, especially with the confirmation of similar interests, based on acceptance of the group presumptions made by the system, is high, especially for marketing purposes. Therefore, given the potential cost savings and information value, the system may provide substantial incentives to the user to cooperate with the information gathering process and to frequently use the system. These incentives may take the form of monetary rewards, coupons, bonuses, contests and random drawings, or improved content or service. These incentives may be allocated and provided in known manner.

Typically, the incentives are allocated according to profit to the proprietor of the database interface system. The greatest opportunities are, indeed, where the user conducts an e-commerce transaction through the portal, wherein the portal is compensated for delivering a willing and able purchaser to a vendor, or wherein the portal itself is the vendor. Therefore, the preferred primary basis for incentives is e-commerce transactions completed. A secondary basis for revenue to the portal includes advertising revenues, typically on a per ad impression or click-through basis. In this case, the proprietor desires mere use of the portal, and primary incentives may be provided, such as a set of useful services, as well as secondary incentives, such as rewards. These services may include, for example, personal shopping or information gathering agents, news feeds, e-mail, personal home page or web sites, electronic wallet services, best price services, consumer review services, on-line auction systems or auction monitoring systems, chat rooms or chat room monitoring services, and the like.

It is well known that by optimizing the presentation of advertising to potential consumers, a higher effectiveness of advertising will be achieved, termed the "ad response rate." It is believed that the probability that a potential consumer will purchase a particular item is correlated with certain personal characteristics, including demographic characteristics, of the person or family unit The field is called demographically targeted advertising. Thus, by predicting the ad response rate for a person, the most highly valued advertisement may be selected for the person. This optimization allows the service provider to charge a higher ad rate, while the advertiser gains more effectiveness for marketing dollars.

Accordingly, one aspect of the invention provides that a conditional probability of a subsequent action by the user may be assessed for each interaction, and that, on the basis of that probability, an economic parameter altered. Thus, for example, the selection of a hyperlink by the user through a browser may be associated with a calculated probability that the user will subsequently purchase a good or service. This probability may then be used to calculate an advertiser charge for delivery of an advertisement, or to prioritize the advertisements sent to the user in order to, for example, maximize the utility to the selected advertiser, the advertisement serving system operator, to the user, or some combination thereof. This calculated probability may also be used to adapt the information subsequently presented to the user. This probability may be calculated, for example, based on a population statistic plus a recent history of the particular user, a collaborative filtering scheme, a long-term monitoring of the user through the use, for example, of cookies and a database, or other scheme, or through express input of user characteristics, such as demographic profile, survey response, or a direct user communication. The logic used to predict the probability may be formal Bayesean, fuzzy logic, a multiple regression equation, neural networks, or other known logic. Further, the probability calculation algorithm need not be completely accurate, so long as it produces an output more accurate than a random selection; however, since an economic valuation is placed on the result, a more accurate calculation will likely be considered more valuable in the marketplace.

It is noted that the advertisement need not be limited to efforts to cause a consumer to purchase. In fact, advertising, as considered herein broadly encompasses seeking to influence a user. Thus, the decisions made by the user need not be purchasing decisions. However, typically, an economic model is appropriate. For example, in a corporate Intranet, messages transmitted to users may be internal messages from within the network, for example informing users of new corporate capabilities, resources, or initiatives, or of changes. According to the present invention, these messages may be delivered in a context-sensitive manner, and based on a user profile. Of course, as in standard consumer advertising, messages may be targeted even to those persons who subjectively resist being so informed, but nevertheless are intended targets of the message. Advantageously, such internal messages may be subjected to an accounting system, wherein a real or imputed economic transfer occurs, for example in the manner of an auction, seeking to maximize the efficiency.

U.S. Pat. No. 6,014,634 (Scroggie, et al.), expressly incorporated herein by reference, relates to a system and method for delivering purchasing incentives and the like to a user, especially using a cookie and associated personal web page.

U.S. Pat. No. 5,974,398 (Hanson, et al.), expressly incorporated herein by reference, relates to a system that allows advertisers to bid for placement in front of particular users based on customer interest profiles.

U.S. Pat. No. 5,933,811 (Angles, et al.), expressly incorporated herein by reference, relates to a system for delivering customized advertisements within interactive communications systems.

U.S. Pat. No. 5,991,735 (Gerace, et al.), expressly incorporated herein by reference, relates to a computerized system for determining a user behavioral profile. This profile may be used, for example, to demographically target advertisements.

Guided Browsing

The use of this information structure therefore presents another particular aspect of the present invention, that of guided browsing. Therefore, the user examines objects using known techniques and systems. The system according to the present invention need not replicate or encompass resource for all such objects. The emphasis of this aspect of the present invention is therefore to facilitate identification of relevant objects through intelligent analysis and information presentation techniques, including, for example, hierarchal or taxonomic organization.

It is noted, however, that the system and method according to the present invention may be integrated with standard object browsing software, such as Microsoft Internet Explorer or Netscape Navigator. For example, custom frames or codes within the command line may invoke particular functions of the present invention. The hierarchal organizational scheme may preempt the standard favorites organization. Of course, the present invention is operable without such integration, and indeed is operable in many instances without a standard object browser at all. In either case, the user is typically given an option to employ standard tools or those enhancements provided by the present invention. Further, many features of the present invention are modular, and need not be employed as a complete set. This is especially the case where features are implemented as sets of small applets, invoked as necessary.

An important trend in the development of the World Wide Web has been the growth of communities, which are web sites organized to encourage communication among groups of people sharing common interests. Such communities have been found to provide an attractive environment for specialized advertising and commercial sponsorship. It can be readily seen that the use of maps and other means for presenting relationships between objects according to the present invention lends itself to the recognition, organization and maintenance of communities. Indeed, the nodal map could be represented somewhat analogously to a demographic map, in which the activity of nodes could be analogized to the size of cities. On an ordinary map, an ordinary village would be represented by a tiny dot, and a large metropolis represented by a larger dot or circle. Analysis of such a map could help provide users an opportunity to initiate or expand a community, or for a service provider with access to such information to stimulate or encourage such a community. Therefore, it is an object of the invention to analyze user profiles, for example taxonomic maps, to define a user's interests, or activities, which may then be used to identify communities which relate to those topics. The user may then subscribe to those communities.

Presentation of Results

After a query is defined, the system may return a large amount of information. Therefore, a proportional burden of information not specifically requested by the user may accompany the download, for example banner ads. Advantageously, however, the user may be provided with options relating to the types and amount of such additional information, and its manner of presentation. Therefore, a variable cost and/or subsidy scheme may be provided.

Once relationships are determined, the output may take any of a number of forms. For example, a tree structure may be created, populated with the available document set. A multidimensional cluster map may also be generated, with trivial dimensions collapsed to give the most useful output image. Thus, discrete, continuous or hybrid techniques may be used for data representation. In a preferred mode, a high level analysis segregates documents based on discrete criteria, such as a taxonomy, although at this level, a single document may be represented in multiple discrete segments. Within each segment, the documents are represented in a continuous map, the presentation of which may be altered by the user as desired to best distinguish the documents of interest.

Using this type of analysis, it is also possible to implement an efficient vector-quantized data compression scheme, based on the common sequences within proximate documents within the hierarchy. Duplicate files would be most efficiently represented. Thus, a series of documents representing a series of drafts of a document may be analyzed to produce a representation of the group as an edit history. This edit history may not only represent the entire set, but in many instances provides a useful organization of the data, including common ancestor documents, draft evolution, and individual contributions.

Digital Rights Management

In fact, some objects according to the present invention include information belonging to thir