Method and apparatus for facilitating use of hypertext links on the world wide web

Abstract

A database server contains pointers to useful information, such as on the World Wide Web. Users of the server may have hypertext links added automatically into documents they submit. Users may additionally contribute to the link database, thereby extending it, and may add additional qualifying information pertaining to the links.

Claims

What is claimed is:

1. A system for organizing competing explanatory information, the system comprising:

hierarchical database means, including hierarchically organized nodes of information;

key-phrase association means, wherein a node of said hierarchical database is associated with a set of one or more key phrases, wherein the key phrases in the set of key phrases share a substantially common meaning;

competing definitions means, wherein a plurality of competing definitions is associated with said set of key phrases, wherein each said competing definition substantially explains the common meaning shared by the key phrases;

property association means, wherein the property association means associates two or more of the plurality of competing definitions with two or more definition properties to produce a set of associated definition properties, wherein at least one of said definition properties is selected from the group consisting of educational level, language, rating, viewer suitability, resource type, context, number of hits, number of installs, and date;

rank-ordering means, wherein the rank ordering means rank orders the plurality of competing definitions based on at least one of said associated definition properties.

2. The system of claim 1, further comprising

search means for specifying a subset of said information in said hierarchical database by means of a keyword-based search;

output means for returning said information subset as a search result.

3. The system of claim 2, wherein said search means further comprises means for specifying a subset of said information in said hierarchical database based on an assigned context for said information.

4. The system of claim 2, wherein said search means further comprises means for specifying a subset of said information in said hierarchical database based on at least one of said definition properties.

5. The system of claim 4, wherein at least two of said associated definition properties correspond to a rating.

6. The system of claim 4, wherein at least one of said associated definition properties corresponds to an owner.

7. The system of claim 1, further comprising

browsing means supporting navigation within said hierarchical database;

output means for returning information corresponding to a current position within said hierarchical database.

8. The system of claim 7, further comprising node selection means based on assigned contexts, and wherein said output means returns information in accordance with said node selection.

9. The system of claim 7, further comprising definition selection means, wherein said competing definitions are selected based on at least one of said associated definition properties, and wherein said output means returns information in accordance with said definition selection.

10. The system of claim 1, wherein at least two of said associated definition properties correspond to a rating.

11. The system of claim 1, wherein at least two of said associated definition properties correspond to an education level.

12. The system of claim 1, wherein paths to nodes of said hierarchical database correspond to semantic contexts.

13. The system of claim 12, further comprising

input means for specifying a set of one or more input key phrases;

ordered search means for searching said semantic contexts in a predetermined order for occurrences of key phrases in said set of input key phrases associated with said nodes of information;

search retrieval means for returning one or more competing definitions from said nodes of information associated with said set of input key phrases.

14. The system of claim 13, wherein said ordered search means further includes means for using a semantic context to limit said searching to a subset of said nodes of information.

15. The system of claim 13, wherein the search retrieval means further includes means for returning one or more competing definitions based on a property list of at least one of said competing definitions.

16. The system of claim 13, wherein the search retrieval means further includes means for returning one or more competing definitions based on an order in which the contexts are ordered by said ordered search means.

17. The system of claim 13, further comprising

link installation means for automatically linking one or more occurrences in a document of an input key phrase to one or more of said competing definitions returned by said search retrieval means.

18. The system of claim 13, wherein the ordered search means terminates upon a first occurrence of a matching key phrase matching any of said input key phrases.

19. The system of claim 18, wherein the search retrieval means returns one or more competing definitions ordered according to one or more properties of said property list.

20. The system of claim 18, wherein the search retrieval means returns a highest ranked competing definition for the matching key phrase.

21. The system of claim 20, further comprising

link installation means, wherein one or more occurrences in a document of an input key phrase is automatically linked based on the highest ranked competing definitions for the matching key phrase.

22. A computer-implemented method for providing a database of competing definitions, the method comprising:

a) storing in a database a hierarchically organized set of nodes;

b) associating with a node in the hierarchically organized set of nodes a set of one or more key phrases, wherein the key phrases in the set of key phrases share a substantially common meaning;

c) receiving and storing a plurality of competing definitions for the node, wherein each of the competing definitions provides a different explanation of the substantially common meaning shared by the key phrases associated with the node;

d) associating properties with the competing definitions to produce a set of associated definition properties, wherein said wherein at least one of said properties is selected from the group consisting of educational level, language, rating, viewer suitability, resource type, context, number of hits, number of installs, and date; and

e) rank ordering the competing definitions using at least one of the associated definition properties.

23. The method of claim 22, wherein said property list includes a property corresponding to an educational level.

24. The method of claim 22, wherein said property list includes a property corresponding to a rating.

25. The method of claim 22, wherein paths to nodes in the hierarchically organized set of nodes correspond to semantic contexts.

26. The method of claim 25, further comprising:

receiving input key phrases;

searching the database nodes in a predetermined order for occurrences of the input key phrases in matching database nodes;

retrieving one or more competing definitions from the matching database nodes.

27. The method of claim 26, wherein the searching is limited based on a semantic context to a subset of said database nodes.

28. The method of claim 26, wherein the retrieving comprises using one or more properties in the property list associated with the competing definitions in the matching database nodes.

29. The method of claim 26, wherein the searching terminates upon a first occurrence of a matching key phrase matching any of said input key phrases.

30. The method of claim 29, further including

linking one or more occurrences of an input key phrase in a document based on the retrieved competing definitions.

31. The method of claim 29, wherein the step of retrieving returns a highest ranked competing definition for the matching key phrase.

32. The method of claim 31, further including

linking one or more occurrences of an input key phrase in a document to the highest ranked competing definition for the matching key phrase.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to facilitating access to information over a computer network such as the Internet. More particularly, the present invention relates to technology for partially automating the linking of documents on the World Wide Web by authors of Web content. Such techniques are particularly useful for more easily creating richly interconnected information on the Web.

2. Description of Related Art

The World Wide Web provides an enormous distributed database of information interconnected physically by the Internet. One of the main difficulties for users of the Web is finding needed information out of the tremendous quantity of information that is available. Various mechanisms have been developed to address this problem.

One mechanism for facilitating access to information on the Web is the index website. An index website is typically a server computer connected to the World Wide Web which maintains an index of Web content that can be searched in various ways by users (clients) connected to the server over the Internet. Indexes are often updated automatically by means of "spiders" which systematically explore the Web looking for new or updated content. Most search engines also provide means for users to install information to be indexed, so that such information may be indexed immediately without waiting for a spider to find it. An example of a premier search engine is the "Alta Vista" website, accessible on the Web at the Universal Resource Locator (URL) address http://www.altavista.com.

A difficulty with search engines is that search results typically contain too much undesired information as well as the desired information. This occurs because the information content of the Web is vast, and because it is difficult for users to construct search parameters in such a way as to pass most desired content while rejecting most undesired content. As a result, users typically must spend a lot of time sifting through search-engine results and/or refining their searches with additional restrictions in the search parameters. Additionally, the information stored in the index is not organized in a form suitable for browsing in a logical order.

Another mechanism developed to facilitate access to information on the World Wide Web is the directory website which presents a hierarchical directory of information that can be browsed by the user. Premier sites of this nature include Yahoo (http://www.yahoo.com), Netscape (http://www.netscape.com), and Excite (http://www.excite.com). A visitor to such a site is first presented with a top-level list of topics. Choosing a topic by clicking on a topic's hypertext link with the mouse produces a list of subtopics, and so on, until a final level is reached at which useful information is displayed about the topic, or else a remote website pertaining to that topic is visited. Directory companies such as Yahoo typically have teams of editors who explore the Web looking for content suitable for reference at their site, and these workers perform a function analogous to the automatic "spiders" used by automated index websites. Like the search engines, directory websites normally support searching within the directory site, thus producing search results of generally higher quality and less "clutter" than typically encountered on an index site. Also like index websites, directory websites typically allow submission of content for reference, subject to editorial consideration. Thus, directory websites improve over index websites by providing editorial selection, logical organization, and browsing capability, all of which are absent in typical index websites.

A first difficulty, however, with directory websites is that they cannot reasonably keep up with the vastness of the information on the World Wide Web by means of manual editorial selection. As a result, directory websites tend to offer far less information relative to index websites. A second difficulty with directory servers is that their content is proprietary and controlled by a team of editors at one company. This editorial control, while ensuring consistently high quality on the site, makes it difficult and sometimes even infeasible for an information provider to obtain a desired listing in the hierarchical directory. One directory site that addresses this difficulty is the Open Directory project (http://dmoz.org/); The Open Directory allows any user on the Internet to become an "editor" for a particular topic at the site. A third difficulty related to the first is that typical directory sites are extremely broad in scope, contributing to the absence of specialized information that is not of interest to a wide general audience.

A difficulty with both index and directory websites is that information is presented without regard to the user's level of education. It is therefore often possible for a high-school senior working on a book report, for example, to encounter information understandable only by a graduate student in a specialized field. There is similarly normally no means for selecting information according to its type or source or other potentially desirable criteria.

To assist users in selecting sources of information, some websites provide a user rating system (or "scoring system") to which any user may contribute. An example of this mechanism is seen in the online book-store website http://www.amazon.com/. Amazon allows any user to contribute a "book review" and an overall rating on a five-star scale. The average rating is displayed for each book, and books which match the user's search criteria are displayed sorted according to decreasing score (and possibly other criteria such as the number sold). An interesting feature of the Amazon rating system is that it is democratic, allowing the vast quantity of World Wide Web users to jointly develop a ranking of the information sources (in this case books). Such a scheme addresses the difficulty of sorting through enormous quantities of information by harnessing a potentially enormous base of users as contributing editors, in effect. A difficulty with rating systems is that they are generally used only at the site where the ratings are collected, and no mechanism is provided for making use of the ratings elsewhere, such as in other documents on the Web linking to the same information.

An important mechanism integral to the function of the World Wide Web is the HyperText Markup Language (HTML) which is a text format supported by Web browser programs (such as Netscape Navigator or Microsoft Internet Explorer). A more recent variant called XML is now gaining support, and its function is similar to that of HTML for present purposes. HTML provides for the specification of hypertext links in Web-page text displayed by the browser. At a minimum, a hypertext link consists of text to be displayed by the browser and a link target which is usually not displayed. For example, the HTML code

<a href="http://www.w3k.org">W3K website</a>

contains the text (also known as the anchor) "w3K website", while the link target is http://www.w3k.org which is a URL pointing to the W3K website. Thus, the link target is normally addressed by a URL pointing to information on the Web about the displayed word or phrase. (The complete HTML format specification may be found online at the URL http://www.w3.org/.) To the browser user, the anchor text of a hypertext link as above appears in a Web-page display as an underlined word or phrase, e.g.,

Visit the W3K website for more information regarding automatic link installation,

and usually in a different color than normal, unlinked text. By clicking on the hypertext link with the mouse, the user directs the browser program to "follow the link" by "navigating" to the URL associated with the link. The link-target URL may point to another Web page anywhere on the World Wide Web, or it may simply point to another location within the same electronic document. Hypertext links in HTML documents make it much easier for the user to explore the World Wide Web by visiting Web pages and clicking on the links found therein. Web browsers further make it easy to return to the page containing the link by using the "back" button, or the "history" list of visited pages maintained by the browser.

A difficulty with hypertext links is that they must be laboriously added by Web content providers. Typical HTML editors merely provide a data-entry form in which the URL for the link target can be typed. A second shortcoming of HTML and Web browsers is that there is no standard mechanism for specifying link properties such as educational level, type of resource, information source, or the like, which could be supported by Web browsers to give the user finer control of link display based on link properties. After the links are typed in, they must be maintained as their URLs change, and as new and better link-targets become available. There is therefore a need for automated assistance with entering, maintaining, and improving hypertext links in documents intended for a hypertext document environment such as the Web.

SUMMARY OF THE INVENTION

It is a primary object of the present invention to facilitate the addition of hypertext links (also called "hyperlinks,""links," or "definitions") to documents intended for access on the Internet via the World Wide Web. Accordingly, the present invention is designed to provide a link installation service which automatically installs hyperlinks within information submitted to the service by hypertext authors. Submissions may be in HTML format, plain ASCII format, LaTeX source format, or a variety of additional formats to be added in the future. The output returned to the user may be in either HTML or LaTeX source format (which may be compiled into HTML format). Criteria can optionally be specified which govern the installation of hyperlinks.

The invention further provides selectable databases of hyperlinks, organized by category (or "context"), which can be optionally selected for automatic link installation. It is further provided that content developers may add their own links to the existing link databases, and they may additionally create new link databases and specify their relation to the existing link databases. Contributing users are preferably required to have a known, verified email address. A user with a verified email address is called a "known user". The invention further provides means for browsing the link databases in a logically organized, hierarchical tree structure, wherein higher-level nodes correspond to more general contexts, and lower-level nodes correspond to more specialized contexts. The link databases can additionally be searched for keyword matches within component fields. Users may provide ratings and/or reviews for individual links in the link databases.

The hyperlink databases of the present invention support various optional "properties" associated with each hyperlink. One such property, useful in the development educational content, is a level designation which indicates the educational level required for best understanding of the link-target information. Additional optional properties include the language of the content (such as English), a viewer suitability rating such as exists for movies (PG-13, R, etc.), and properties defined by the user. Link properties can be specified by users to control the automatic installation of links, and/or to control what is displayed while browsing the link databases.

Educational levels not specified on submission are estimated based on the level of links found within the link target document. As a result, every link in the link database is assigned either an educational level, either manually or automatically. Determining levels automatically detects any "cycles" in the link database. (A "cycle" occurs when document A links either directly or indirectly to document B, and document B links either directly or indirectly to document A.) Cycle detection can help content providers eliminate inadvertent "forward references." Means are provided for marking forward-reference links in submitted documents so that educational level will not be affected. Cycle-free systems of links can be more effectively used as a basis for online course materials.

Another feature of the present invention is the ability for users to rate (or score) the quality of any link in the database and/or to submit a written review of any link. The quality ratings may be averaged together and used to determine the relative ordering of the links when there are multiple link targets for the same word or phrase ("competing definitions"). In the typical case of HTML format, features of the JavaScript scripting language may be used to provide convenient access to multiple link targets, ranked according to score. Alternatively, the latest ranked list of competing definitions may be maintained on a central server on the Web, with the installed link pointing there, instead of containing only a snapshot at the time of link installation, which may rapidly go out of date. Alternatively, the currently highest rated link may be installed in the user's Web document for each recognized topic.

BRIEF DESCRIPTION OF THE DRAWING FIGURES

FIG. 1 shows an example initial Web page seen by a visitor using a Web browser to access the online version of the service.

FIG. 2 shows a Web page giving an overview of the capabilities of the online service.

FIG. 3 shows an example Web top-level page seen while browsing the hyperlink databases.

FIG. 4 shows an example lower-level page seen while browsing the hyperlink databases, in which the context has been narrowed considerably.

FIG. 5 shows an example browsing view at the level of a key phrase in which all displayed links are interpreted as "definitions" for the key phrase.

FIG. 6 shows an example form for adding a new link (definition) to the link database for the current key phrase.

FIG. 7 shows a Web page for submitting text for link installation.

FIG. 8 depicts the tree structure of the hierarchical link database.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The following is a description of the best presently contemplated modes of carrying out the invention. The descriptions are not to be taken in a limiting sense but are made for the purpose of illustrating the general principles of the invention. It is particularly noted that the invention may be implemented in a variety of different file formats, database technologies, search and replace methods, computer processors and system architectures, host operating systems, network protocols, user-interface frameworks, and the like.

Client-Server Architecture on the World Wide Web

FIG. 1 illustrates how a World Wide Web "home page" might appear on a website embodying the principles of the present invention. The user has several choices of where to "navigate" next:

The first choice 101 is a hypertext link entitled "Learn about the W3K," where in this example, "W3K" is an acronym standing for the "World Wide Web of Knowledge." If this choice is selected by clicking the mouse on the underlined text, the visitor "navigates" to the Web page shown in FIG. 2 which provides an overview of the online service provided by the W3K. In particular, it is explained how submitting plain text 110 to the W3K server will result in hyperlinked text 111 being returned to the user. A summary 115 of high-level functions is also provided in FIG. 2.

The second choice in FIG. 1 is a hypertext link 102 entitled "Browse the W3K." If this choice is selected by clicking the mouse on the underlined text, the visitor "navigates" to the Web page shown in FIG. 3 supporting browsing of the hyperlink databases, as described further below.

The third choice in FIG. 1 is a hypertext link 103 in which the text displayed by the Web browser is "Install W3K links in a Web document." If this choice is selected, the visitor is taken to the Web page of FIG. 7 where the user can submit text for link installation in a variety of formats. The text is returned to the user by the server with hypertext links installed according to the user's specifications. Link databases to be searched can be collected into a list during the browsing operation.

The fourth choice, "Add to or Edit the W3K," is a link 104 to a Web page for editing the link databases. Editing operations include including submitting new links, creating new link categories, and changing previously submitted links or link properties. These editing functions are also available while browsing the databases.

The fifth and final choice, "Search the W3K Dictionaries," is a link 105 to a Web page for specifying search criteria in terms of link properties. The search collects together all links in the link databases matching the search criteria, and displays them organized by properties according to user specifications. The search feature is useful for collecting various link subsets together for various purposes including link installation, editing link properties, and other functions involving groups of links. As an alternative to a list display format, a sparse context hierarchy can be generated, containing only the database information matching the search criteria; the sparse hierarchy can then be conveniently browsed by the user.

These functions are described in further detail below.

Overview and Terminology

This section introduces the main terms which will be used hereafter.

Hierarchical Contexts

The link databases are organized hierarchically according to category, somewhat like the Dewey decimal system for library organization. Each category (or "directory") is interpreted as a context analogous to a field of study. Each context may itself contain any number of contexts ("subcontexts," or "subdirectories"), and it may additionally contain a database of information pertaining to that context (which may be a implemented in a file in that directory).

The particular sequence of directories obtained by visiting one subdirectory after another is called a path. Every context may be identified by the directory path that reaches it from the top level. Thus, the set of all contexts form a "tree structure" analogous to the hierarchical file systems used by all major computer operating systems at the present time.

Dictionaries

A link database (or "dictionary") preferably comprises a list of (key,URL) pairs. A key (or "key phrase" or sometimes "word") identifies a topic or concept, and the URL points to information about that topic on the Internet. In a loose analogy with an ordinary dictionary, the key is the "word being looked up", and the URL points to its "definition". However, unlike an ordinary dictionary, the (key,URL) pairs in the link database are interpreted within the particular context associated with the directory containing that dictionary. In a somewhat better analogy with a technical encyclopedia in a particular field, the key corresponds to the noun phrase identifying a technical topic for which an article exists in the encyclopedia, the URL may correspond to the page number on which the article begins, and the context may correspond to the technical field for which the encyclopedia was written.

Because dictionaries are interpreted in a particular context, alternate definitions are not allowed. In other words, a context is preferably sufficiently narrow such that all terms (words or key phrases) in that context have a unique meaning. Ordinary "flat" dictionaries must accommodate alternate definitions for a single word, while "hierarchical dictionaries" need not. Thus, if a term is found to have a second meaning in a particular context, it is time to create one or more subcontexts in which that term is disambiguated.

Synonyms

A single URL can provide only one "definition". However, a single URL can be used to "define" any number of key phrases, which are then regarded as synonyms. Often the title of the addressed HTML page on the Web is the "key phrase" that is "defined" by the URL. When there are several (key,URL) pairs having the same URL, the different keys are treated as alternate phrasings for the same concept or topic, and are said for form a synonym group. The following example dictionary entries provide an example of a synonym group:

KEY=Taylor Series Expansion

URL=http://www.mathworld.org/analysis/TaylorSeries.html

KEY=Taylor Expansion

URL=http://www.mathworld.org/analysis/TaylorSeries.html

KEY=Taylor Series

URL=http://www.mathworld.org/analysis/TaylorSeries.html

Order is important in the dictionary because "the first match wins" during automatic link installation. For example, with the above ordering, the key phrase "Taylor Series Expansion" will match before checking for "Taylor Expansion" or "Taylor Series". Ordering equivalent key phrases from longest to shortest ensures that the longest possible match will occur in documents submitted for link installation.

Synonyms can be listed in a link's properties, or they can simply be entered as additional link entries pointing to the same link target (URL), since links take up relatively little space.

"Key Phrase" Directories

A "key phrase" may be understood as a bottom-level subdirectory of the context tree. A key-phrase directory holds a dictionary (link database) containing at least one link. This database may be implemented as a file residing in a directory having a name derived from the key phrase. Preferably, however, all key phrases in a particular context (together with their links), plus perhaps additional contexts, are implemented in a single larger database file in the parent context directory. For simplicity, however, a key phrase will nevertheless be considered logically to be a bottom-level directory (leaf node) in the hierarchical context directory, irrespective of implementation details associated with the use of a hierarchical file system.

All of the links in the key-phrase directory are interpreted as competing sources of information on the one topic identified by the key phrase. The tangible difference between a key-phrase directory and a context directory is that the key-phrase directory has no subcontexts, only links. Thus, a bottom-level directory in the context tree hierarchy (a "leaf node" of the context tree) corresponds to a single concept or topic, and all of the (key,URL) pairs in its dictionary pertain to that one topic. The number of distinct URLs present is the number of competing sources of information.

Perhaps the simplest means for handling synonyms is to add a key-phrase directory for each alternate phrasing of each topic. Because order is important when traversing a synonym group, the context-tree implementation must provide a means for ordering sub-directories, at least when those sub-directories correspond to key phrases. Alternatively, an ordered link database file may reside in the context directory containing the key phrase, and include all other key phrases in that context as well (including synonyms); the key phrase(s) corresponding to each link may be stored as link properties.

In the preferred embodiment, synonyms are not handled as separate key-phrase directories. Instead, a single representative is selected (usually the most descriptive or canonical), and all other equivalent phrasings (synonyms) are listed in a separate synonym file in the key-phrase directory. (Order is carefully preserved.) During browsing, synonyms are displayed at the bottom of the key-phrase page.

Context Synonyms

The preferred embodiment also supports context synonyms, as opposed to key-phrase synonyms just described. Context synonyms are presently implemented using symbolic links in a UNIX file system implementation of the context tree. As an example, the context hierarchy below illustrates two different paths to the subcontext (directory) "Sound_Synthesis", where the notation "-->" indicates a symbolic link, as is typically done when listing files in a UNIX file system:

Humanities

Music

Computer_Music

Sound_Synthesis

Engineering

Electrical

Signal_Processing

Sound_Synthesis -->/Humanities/Music/Computer.sub.13 Music/Sound_Synthesis

In this example, the "true parent" of the node Sound_Synthesis is Computer_Music, while the parent Signal_Processing is a "linked parent". There can be any number of linked parents, but only one true parent.

Symbolic links provide a means for reaching multidisciplinary fields by browsing the constituent fields in a top-down way. At any time, a symbolic link may be deleted and replaced with a copy of some or all of the directory which was formerly linked (possibly utilizing symbolic links at a lower level). In this way, closely related contexts may start out as identical, but later may evolve into separate collections, as the maintainers see fit.

Context Dictionaries

The dictionary corresponding to a particular context is defined as the union of all key-phrase dictionaries in that context. In other words, the dictionaries belonging to all key-phrase subcategories in the particular context are logically concatenated together into one large dictionary, with additional dictionary entries added for all synonyms. This dictionary is referred to as the "context dictionary". The order in which the key phrases are concatenated is prescribed.

When installing links in submitted documents, several context dictionaries are normally combined together to form a larger "aggregate dictionary" which is what is used for link installation. A typical aggregate dictionary consists of the context dictionary for the "current context" (established, e.g., by browsing), followed by the context-dictionaries of all subcontexts (usually not in any particular order, unless explicitly listed by the user), then followed by the context dictionary of the "true parent" context, followed by the context dictionary of the true parent's true parent, and so on, until the context dictionary of the top-level directory is appended (which contains extremely generic terms). Linked parent dictionaries may also be added in where desired. Since order is respected during link installation, definitions provided in the "current context" will receive first precedence, followed by definitions occurring in subcontexts (which are considered within the current context), followed by the more generic definitions of parent contexts. Since "the first match wins" in link installation, generic terms defined in parent contexts are "overridden" by more specialized definitions of the same terms of art in the current context. For example, the word "resolution" might be defined at the top level as the first definition appearing in an ordinary dictionary of the English language, while in the context of " . . . /Signal_Processing/Spectrum_Analysis" it would be given its more arcane definition regarding the resolving power of a short-time Fourier transform.

Browsing the Link Databases

There several benefits to providing browsing of the link databases:

it provides a unique educational resource which organizes valuable information on the Web in a manner especially well suited for educational purposes;

it provides a convenient means for learning what links are available for installation in documents;

it provides a convenient means for collecting context dictionaries for subsequent use in automatic link installation. While browsing, links and/or entire context subtrees can be marked for inclusion or exclusion in subsequent automatic link installations;

it provides a convenient means for navigating to contexts in which subcontexts and/or key-phrases can be added and/or edited by the user, or to key-phrase directories in which links can be added and/or edited and/or rated; and

by displaying links selectively according various link properties, browsing provides a means for viewing useful link subsets, such as all links entered by the user or user's group.

Link database browsing support on the server may be implemented in a variety of ways. As an example, there are commercially available scripts which implement directory websites, such as the links-2.0 scripts from Gossamer Threads, Inc., and such scripts can be adapted to implement the hierarchical dictionary of the present invention. FIGS. 3 and 5 illustrate the appearance of such a browsing system. Alternatively, one may use HTML SELECT pop-up menus, which are dynamically generated from the current directory contents. However, for performance reasons, static HTML pages are preferable over dynamic HTML generated by the server, when feasible. To provide more context and ease of navigation, the database directory structure may additionally be displayed in a fixed HTML frame on the left, as is currently done on many websites. For example, the way directory trees are displayed on the left in Microsoft Windows Explorer is a good model.

FIG. 3 illustrates a Web page display at the start of browsing. The top-level context is displayed. In this simplified example, only four top-level subcontexts are offered (Computing 131, Education 132, Legal 133, and Music 134). Each of these words is a hypertext link, which can be clicked with the mouse to navigate to the associated sub-context. For example, clicking on Music 134, then on "Computer Music" (which is available in the Music context), then on "Signal Processing", then finally on "Sound Synthesis" produces the page shown in FIG. 4.

The Standard Browsing Menu

Near the top of each page during browsing is a set of hypertext links 130 separated by a vertical bar `.vertline.`. This is the "standard menu" appearing at the top of every page while browsing the W3K website and at other times as well. Each of these links allows the user to carry out some available function.

The "W3K Home" link in the standard menu 130 takes the browser back to the initial W3K home page illustrated in FIG. 1, as does clicking on the W3K logo.

The "Browse from Top" link in the standard menu 130 navigates to the top-level browsing page shown in FIG. 3.

The "Select Hierarchy" link navigates to a page where a different context hierarchy can be selected for browsing. There is only one primary public context hierarchy (the one reached from the second choice 102 in FIG. 1). However, individual users and groups of users can set up context hierarchies for their own purposes, without having to worry about fitting into the ever-expanding primary public context hierarchy. If well known "language localization" methods are not available, as preferred, to provide alternate language selection for each Web page in the public hierarchy, alternate hierarchies can be used to support alternate languages. Alternate hierarchies can be designated by their creators as public (anyone can add to it), restricted (anyone can read it, but only the owner(s) can write it), or private (only the owner(s) can read or write it). The owners include the creator and members of any groups listed by the creator as being co-owners.

The "Install Links" link in the standard menu 130 navigates to the form provided for submitting documents for link installation, which will be described further below.

The "Add Subcontext" link navigates to the form provided for creating a new subcontext within the current context. Since FIG. 3 is at the top-level context, this operation is only allowed in a public hierarchy for a select group of "trusted" users.

The "Add Key-Phrase" link leads to the form for adding a new key phrase in the current context. At least one definition link is required when adding a new key phrase. At the top level of the public hierarchy, this operation is restricted to trusted users since any key phrases appearing at the top of the hierarchy are "generic terms" having definitions which are independent of context. Truly global key phrases such as domain names and trademarks are appropriate at the top level.

The "Add Definition" link is for adding a new definition for a key phrase. This entails supplying a URL which points to information about the URL and some other information, as will be later described. Thus, the number of distinct URLs in the set of URLs associated with a particular key phrase can be increased from 1 (its usual initial value) to any number by adding more definitions. The "Add Synonym" link in the standard menu 130 allows the addition of a key phrase to a list of "synonyms" for an existing key phrase. A synonym can also be constructed for a context. Synonyms will be described further below.

"Submit Dictionary File" provides convenient submission of a large numbers of links (key phrases and definitions) as well as the ability to specify context path for each one, as will be described. It is additionally possible to display specified contexts and contexts selected for link installation in the form of a dictionary file. For example, a user can perform a search in order to collect all links contributed by that user, display the results as a dictionary file, save the dictionary file on his or her local computer, perform any desired editing operations, and submit the edited dictionary file back to the server to update his or her links on the server.

"Modify Additions" allows the user to edit (modify or delete) any information he or she submitted to the W3K site. In particular, it is possible to modify link properties, delete a link, delete a context or key-phrase directory wholly owned by the user or user's group, and so on. A user belonging to one or more groups may edit any information submitted by anyone in any of those groups. A set of records to be edited can be created by means of the search facility. A record may hold the information associated with a link, key-phrase directory, or context directory.

"Select Context" selects the "dictionary" associated with the current context for inclusion in subsequent "link installation". The context dictionary normally includes each key phrase in the current context together with at least one definition for each key phrase. It may also include similar information from parent contexts and subcontexts, as will be discussed. Thus, the aggregate dictionary used in link installation is like a kind of "shopping cart" that can be filled with component dictionaries found while browsing around the context hierarchy; in this analogy, "items to be purchased" correspond to the dictionaries to be used in link installation.

It is also possible to assemble various context directories into an aggregate dictionary for link installation without browsing by simply providing a context dictionary list, or by selecting contexts from a number of SELECT menus in HTML listing all available contexts. After the current context is selected, the "Select Context" link changes to "Deselect Context", so that clicking on it takes the current context out of the aggregate link installation dictionary.

Selection configuration information lower in the hierarchy is not modified when excluding a context, so that re-selecting the context allows the contained selection configuration to become active once again. During link-installation (FIG. 7), it is possible to override all such selection information by simply specifying an explicit list 75 of context dictionaries, or selecting "All W3K contexts" in the form entry for contexts 179.

Browser "cookies" are very useful for storing the context search preferences for the user across sessions; since many tend to work in one or a few fields, it is often the case that the contexts used for link installation do not change very often. Browser cookies are simply information stored on the user's computer (the client computer) by the server; cookie files are supported by the major Web browsers such as Netscape Navigator and Microsoft Internet Explorer. If cookie files are not available for any reason (they can be disabled by the user), preference information can be stored on the server indexed by the user's email address, which is unique among users.

"Browse All Selected" places hierarchy browsing in a special mode in which only the currently Selected contexts and links are visible. This can also be reversed so that only deselected contexts are visible. (Sometimes it is helpful to go back and forth.) This feature can help the user more quickly review what link databases ("key-phrases" and "definitions") have been selected for link installation.

"Edit All Selected" is similar to "Modify Additions" except that instead of determining the list of database elements to be edited by using a search (or direct specification), it is initialized from the set of selected links owned by the user and/or groups to which the user belongs.

"What's New" creates a list of all contexts, key phrases, or definitions which have been added recently to the system.

"What's Cool" creates a list of all contexts, key phrases, or definitions which have been receiving relatively high traffic ("hits") recently.

"Top Rated" creates a list of highest ranked links in the database. These are generally excellent home pages, tutorials and the like on various topics.

"Email Updates" allows the user to subscribe to the W3K newsletter.

"Random Link" takes the user to a randomly chosen definition link.

"Search" supports general search for information within the current context and beyond.

Context Path Display

While browsing, the "context path"140 (FIG. 4) is displayed just below the standard menu 130, with each path element separated by a colon `:`. In FIG. 4, for example, the context path is displayed as "Top: Music: Computer Music: Signal Processing: Sound Synthesis." Clicking on the "Lagrange Interpolation" key-phrase 144 in this context takes the browser to the definition page for Lagrange Interpolation shown in FIG. 5.

Search Form

Below the horizontal line in FIG. 3 is a search form. Typing text into the field and clicking on the "Search!" button results in a dynamically generated web page listing all links (in all contexts) matching the search criteria. More refined searches can be carried out by first selecting the "More search options" link. Since links have quite a few properties (to be discussed), searches can be honed rather finely without relying entirely on typical means for selecting a subset of all names and phrases within contexts, key-phrases, and definitions.

Topics under a Context

FIG. 4 displays the contents of the context-path

/Music/Computer_Music/Signal_Processing/Sound_Synthesis.

We see that the "Sound.sub.13 Synthesis" context contains two subcontexts "Acoustic Instruments" 148 and "Vintage Methods" 149.

In addition to subcontexts, there is a list labeled "Words and phrases defined in context Sound Synthesis" 141. (For greater convenience when browsing contexts, browsing can be configured to show only a single link to the key-phrase list on a separate page.) The phrases listed include "Commuted Synthesis" 142, "Physical Modeling" 145, and "Lagrange Interpolation" 144.

Technically, as far as the browsing function is concerned, "words and phrases" (key phrases) are similar to "subcontexts". However, key phrases are browser categories with no subcategories, only links, while contexts are browser categories containing subcategories (either subcontexts or key phrases). The links under a key-phrase are treated as "competing definitions" for that key phrase.

FIG. 8 illustrates the relationships among contexts, key phrases, and definitions. The top level context 180 is the root node of the tree structure defined by the hierarchical link database. There can be any number of subcontexts or key phrases under the top level context 180. In the example of FIG. 8, there are two subcontexts, "Intermediate Context 1" 181 and "Intermediate Context 2" 182. Since these are context directories, they each may contain any number of subcontexts and/or key phrases. In the present example, there are two key phrases 183 and 184 in the first subcontext 181 and one synonym group 185 (two equivalent key phrases) in the second context 182. A key phrase must have at least one definition (link) associated with it. In the present example, "Key Phrase 1" 183 contains three competing definitions 186, "Key Phrase 2" 184 contains four competing definitions 187, while the synonym group 185 consisting of "Key Phrase 3" and "Key Phrase 4" contains two definitions 188 to choose from for that synonym group. Since order is important, we may choose a consistent ordering convention for tree diagrams in which the ordering of all subnodes of a node is defined as left to right in a diagram as in FIG. 8.

Note also in FIG. 4 that the "Sound Synthesis" 141 context includes one synonym 143. This is a context synonym identified by the path

Engineering: Signal Processing: Sound Synthesis

which can be thought of as a different context path to the same place. A context synonym can be thought of as a "symbolic link," in the sense of a UNIX file system, from one "context directory" to another. It is often appropriate for multidisciplinary fields, such as the field of sound synthesis, which belong as a subcontext of more than one high-level context. In link installation, context synonyms can provide what is analogous in computer science as "multiple inheritance", i.e., the dictionaries of multiple parents ("Music" and "Engineering" in this example) can optionally be included automatically in the formation of the aggregate dictionary for link installation, while only the one main context ("Sound Synthesis" in this example) has to be selected for link installation.

To illustrate a "context synonym in FIG. 8, we could add a third subcontext box under the "Top Level Context" box 180 entitled "Intermediate Context 3" which could have a different kind of border to indicate that it is a symbolic link to some other context. We could then draw an arrow from the "Intermediate Context 3" box to its equivalent, such as either "Intermediate Context 1" 181 or "Intermediate Context 2" 182.

FIG. 5 shows a display of two "competing definitions" for the phrase "Lagrange Interpolation" 151. Either of the two links 152 or 153 may be installed in a document containing the phrase "Lagrange Interpolation". They are both named "Lagrange_Interpolation" because that happens to be the title of both Web documents. However, the links point to two different targets on the Web written by two different authors.

The "new" superscript after a link 152 or 153 means it was added relatively recently. In this example, both links for "Lagrange Interpolation" were added on the same day.

The "popular" superscipt for a link 152 or 153 means it has been receiving relatively frequent visits (or "hits") via the W3K site. The number of hits displayed in this case is 0.

Also displayed in FIG. 5 for each link 152 or 153 is the date 155 the link was submitted, the number of hits 156 (number of times a anyone as clicked on the link at this site), a rating 157 for each link (which is 0 since the links were just added), and the number of votes included in each rating (also 0 at the moment). Available elsewhere on the website also is the number of times a link has been installed in Web documents. Finally, there are three links 154 for rating each link (assigning a quality score from 1 to 10 and optionally submitting a more detailed written review), reading the reviews written by others, and viewing all of the link's properties in tabular form. After the rating display is a hyperlink which a user can select in order to contribute a rating or a review of the link.

This completes a first-pass overview of the main pages and selections seen by the user while browsing the link databases. Functions available while browsing will be described further in the following sections.

Adding or Modifying Definitions or Categories

In FIG. 5, the "Add a Definition" link 158 navigates to the form shown in FIG. 6 for adding another definition link for Lagrange Interpolation. The current key phrase "Lagrange Interpolation" is filled into the "Topic" field 161, and the context path leading to the key phrase is filled into to "Context" field 160. This makes it convenient to enter a new source of information (definition) on a topic (key phrase) while browsing.

When "Add a Category" or "Add a Key Phrase" is selected from the top-level context (or "Add . . . " is selected on the main website home page), the "current context" field of the form becomes instead a pop-up HTML "select" list containing all of the contexts presently in the database, making it convenient to quickly select any context in which a new subcontext or key-phrase is to be added.

The only required fields on the add-definition form (FIG. 5) are the URL 162 and user's email address 169. All others are optional.

The URL is the new definition, and it is tested by the server to make sure it is responding. If the Site Title field 163 was left blank, the title of the Web page addressed by the URL, which is automatically retrieved by the server (using the Perl LWP module), is filled in automatically as the link title.

The contributor's email address is required because all submissions to the server in the preferred embodiment are associated with the contributor's email address. However, there are alternative means for identifying users known in the art, such as a more conventional registration procedure in which the user chooses a login name and password. The preferred embodiment ensures that the email address given really reaches the user. If the user is new, an authorization process, described in .sctn.5.1.9, is initiated which tests the user's email address.

While not required, the link contributor is invited to write a short description 164 of the website, specify the minimum 165 and maximum 166 educational level covered at the site (usually done by the author of the site), and specify the type of resource 167 (home page, conference paper, book chapter, or the like). The user may also type in his or her name 168.

Fields such as educational level 165 that are potentially confusing tend to have a "Help" link 62 next to them. For example, the educational level help 62 explains that the numerical value is in units (loosely) of "years of education likely required to understand the material". A minimum level with no maximum level corresponds to setting one level rather than a range of levels. When no educational level at all is provided with the definition, the link server will attempt to compute it automatically based on the level of the links it contains, as will be described. In a script-based submission, finer control is possible using additional level-related properties.

Things like "educational level" and "resource type" are examples of link properties. The context path leading to a link is also one of its properties, as is its URL, title, description, and so on. A link can have more properties than these, some of which will be described below. The "Specify Additional Properties" link 61 takes the user to a larger form where the additional properties can be specified.

When the user is satisfied with the filled in definition-submission form, the submit button 60 can be pressed to send the form to the link database server (a computer at w3k.org in this case). At that point, the server tests the URL by retrieving the first page, checks that the user's email. address is known and that the user's IP address and cookie information match information previously stored on the server (otherwise authorization is carried out), checks for duplication of the key phrase and URL in the given context, possibly checks the URL target for "inappropriate content", assigns an automatic educational level if none was provided (unless automatic level assignment is already scheduled at regular intervals), and adds the new definition to the link database for the current key phrase (and context path, if the database file holds links for multiple key-phrase contexts). If the addition was successful, the user is navigated to a dynamically generated Web page summarizing the information added to the database. If there were any problems, an error page is generated listing the reason(s) for failure to accept the page.

A far quicker means of entering definitions is by means of dictionary file submission which can be regarded as a script-based replacement of the above browser-based interface. An example of such a dictionary file is given in a later section. The form for submitting such a file may be reached via the "Submit Dictionary File" link in the standard menu, or as an option under the "Add to or Edit the W3K" option on the server home page.

Private Context Trees

As mentioned when describing the standard menu 130, known users may optionally create a new top-level context tree which is private to that user or to one or more groups identified by the user. This mode of usage is advantageous for private usage without incurring collisions with links in the main "global" context tree shown in FIG. 3. It is further the only way a known user can submit large quantities of contexts, key phrases, and links by means of a dictionary file submission, since that operation is not permitted in the global public context hierarchy. Further details will be described.

UserAuthorzation

Whenever a user requests an operation on the server requiring information to be stored on the server (any "editing operation"), the user must be "known." Being known means the email address of the user has been given by the user to the server, and the email address has been verified by the server to work (reach the user). When an editing operation of any kind is requested (including the simplest form of link submission, or even a link rating from 1 to 10), if the user is not yet known, an "authorization process" is carried out as a preliminary step in the desired editing operation.

In the authorization process, the user submits his or her email address in a simple Web-page form, and the server (1) emails a randomly generated ASCII string to that email address, and (2) navigates the user to a Web page containing a form for receiving that random string from the user. The form also instructs the user to receive the email and to paste the random string into the second authorization form and submit it. This process verifies that the email address in fact reaches the user.

The email address and IP address of the user are then saved on the server. Additionally, the same information is written on the user's computer using a browser cookie. If the cookie goes away for any reason, or if the user later comes in from a different IP address for which authorization has never occurred (e.g., due receiving a new dynamically assigned IP address from an ISP, or using for the first time a different home computer connected directly to the Internet), authorization is triggered once again when any editing operation is requested. Users coming in over dynamically assigned IP address generally have to be authorized for each session until all such IP addresses have been seen and logged on the server along with the user's email address.

After a successful authorization, the user may use the "Back" button in his or her Web browser to find the page which triggered the authorization process, and resubmit the form successfully.

Link Properties

Many other properties can be specified for a link besides the URL 162 and email address 169. One of the most important properties, brought out in the main form, is educational level 165. Both a minimum 165 and maximum level 166 can be set. When the link-target document is written at a single well-defined educational level, such as ".sub.10 th grade", the min and max can be set to the same value (such as 10), or the max can be left unset (which defaults to level 100, meaning no maximum). When the document spans a wide range of educational levels, such as a well designed "topic home page" might do, the min and max can be set appropriately to cover the estimated range. The minimum level still sets the official "level" used in automatic level assignment for other documents, but the maximum level, if specified, may affect link installation when a specific level range is specified for that. An educational level is implemented a floating-point number, so that a level of 10.5 can be specified, e.g., in the form 165 or 166.

Another important link property, also on the main form, is resource type 167. Resource types include dictionary definition, encyclopedia article, unpublished article, conference paper, talk overheads, refereed journal article, book chapter, book, tutorial, lecture notes, course readers, and the like. Sometimes authors may wish to screen out non-refereed sources such as conference papers or unpublished works. Of course, refereed publications and books will typically be hosted on the website of a publisher, requiring some form of payment for access, such as a site subscription or, preferably, a per-page "micropayment" such as the well known Millicent system provides.

Additional optional properties may be specified on a second form by selecting the "Specify Additional Properties" link 61. Additional properties include source type (individual, educational institution, company, non-profit organization, etc.), geographical location, language (English is assumed by default), "viewer suitability" analogous to `PG-13`, `R`, etc., for movies, a list of groups to be granted editing access, and so on.

Link properties added automatically by the server when installing a link in a database include a unique integer ID, the email address and IP address of the link contributor, the date of submission, an initial rating of zero, an initial zero number of "hits", an initial zero number of "installs" in documents, and the like.

Link properties make it convenient to specify "virtual link database directories" which include only the links satisfying certain criteria specifiable in terms of link properties. For example, a user may ask to see only tutorials and books in a certain educational level range. Alternatively, an author may specify seeing only links belonging to that author's email address, or group. Thus, properties enable selective browsing (or listing) as well as more selective link installation. Such selective browsing may be specified using the Search feature on the site home page 105, standard menu 130, or at the bottom of any Web page seen while browsing.

Link properties may also be usefully included in installed links (within HTML "comments" or in specially defined XML tags) when indirect links are being installed (that is, when the installed link points to a centralized link server which forwards the user's browser to the ultimate destination). Installed link properties may be interpreted by the link server to provide additional control over link behavior. For example, a teacher using Internet documents for a 9.sup.th grade class could configure the link server to suppress all links having an educational level greater than 10. That way, when educational level properties are available for all links, as the present invention provides, documents may be populated with hyperlinks which can be configured not to refer a student to information at a more advanced level than the teacher desires. The teacher may further suppress any links with a viewer suitability rating below a certain value. In summary, installed link properties enable dynamically configurable link behavior based on link property values.

In another use of installed link properties (which requires either browser support and/or local editing of the HTML containing the installed links), link properties can be associated with "classes" in "cascading style sheets" (an add-on to HTML) in order to display links to dictionary definitions in one manner, encyclopedia articles in another manner, and home pages in another, etc.

Restricted Directories

When a subcontext is created, it can be marked as "restricted" to the owner (creating user) or to groups specified by the owner. Restricting a directory prevents anyone but the owner or specified groups from modifying the subdirectory. The restricted directory can optionally be made "invisible" to users other than those having modification rights, in which case the restricted directory is said to be "private". An unrestricted directory is said to be "public". A restricted directory can be deleted or renamed or otherwise reorganized no matter what it contains. Typical uses of restricted directories include

Retaining the ability delete the entire directory and rebuild it with a dictionary file submission.

Supporting a private dictionary corresponding to a particular project, such as a book, in which it is desired to have complete control over all links used in link installation.

The name of a restricted directory has the name of its first group (or owner, if no access groups are defined) automatically appended as a suffix to the name chosen by the owner in order to prevent conflicts with public directories and other restricted directories on the same topic. With this convention, any number of users may have restricted subdirectories on the same topic. For example, in the subdirectory "/Music/Computer_Music/Synthesis/" there could be

Commuted_Synthesis_by_mak@vipunen.hut.fi/

Commuted_Synthesis_by_jos@ccrma.stanford.edu/

In this way, any number of experts may provide their own "packages" of links on the same topic.

A known user may even create a new top-level hierarchy which may be designated public, restricted, or private. User- or group-owned hierarchies of this nature which lie outside the primary public hierarchy may be placed in a special standard menu item entitled "Alternate Universes", e.g., to indicate that they are not a part of the primary public context hierarchy.

Link Ratings and Reviews

When browsing reaches a key-phrase directory, as shown in FIG. 5, following each competing definition 152 or 153 is the hyperlink "Rate It" which navigates to a form where that definition (link) can be rated on a scale from 1 to 10, and/or a written review about that link can be submitted. If the user is not known, an attempt to submit a rating or review routes the user to the authorization page, and after a successful authorization, the rating or review is accepted by the server.

All ratings and reviews are stored on the server along with the email address (and IP address) of the contributor. Only one rating and review are allowed per item per email address, but the user owning the rating or review can modify either at any time. Certain "trusted" users, such as website editors or expert consultants enlisted to help with ratings and reviews, may be given higher weighting in the ratings, and the reviews may be organized by editors according to their quality. Otherwise, the rating system is straightforward and similar in functionality to the five-star rating and review system used at http://www.amazon.com for books.

Link Installation

A primary function of the invention is to facilitate the installation of hyperlinks in documents intended for the World Wide Web. This section provides a detailed description of link installation in the preferred embodiment.

Installed-Link Types

There are at least four alternative ways to install a link in a document.

In the first mode, a hypertext link is installed directly to the top-ranked source of information on the topic identified by the matching key phrase in the user's submitted text. This is the first choice presented in the "Link Type" radio-button-group 177 of the default link installation form (FIG. 7). A disadvantage of this approach is that links often become "stale" due to changing ISPs, changing filenames, etc., requiring the links to be re-installed from time to time. (The link installation server preferably tests all links in its databases periodically and eliminates them if they are unavailable for a prolonged period of time, such as more than a week. When all links containing a bad URL are automatically removed from the databases, all owners of the links are notified automatically by email and invited to submit an updated version of the link(s).)

The second approach is to install an indirect link which links via a centralized server (such as a website providing the link installation service). This choice is provided by the second radio button in the "Link Type" portion 177 of the default link installation form. Such an intermediate website acts as a so-called "proxy server" for the link. Indirect links may always point to the most up-to-date, top-ranked source of information on any given topic. An example URL syntax for this mode of operation is

http://www.w3k.org/jump.cgi?ID=35 where it is assumed that each link has a unique integer identifier on the proxy server, and jump. cgi is a CGI script which is passed the identifier as if it were a form submission in which the form contained a field named "ID" with the value 35. To avoid having to assign unique identifiers across all contexts, the context path can be included in the URL, e.g.,

http://www.w3k.org/jump.cgi?ID=3&PATH=Engineering+Signal_Processing

Context paths can similarly be assigned integer IDs in order to shorten indirect URLs.

A third approach is to insert a link to the "key-phrase page" itself at the centralized server (the page on the server listing all "competing definitions" for that key phrase). This is the third and final choice in the "Link Type" radio group 177. In this case, an end user following such an installed link will see all competing definitions, in ranked order, instead of only one. The end user can then request that the definitions be reorganized according to various criteria such as educational level, document size, type of resource (article, book, etc.), type of source (.edu, .org, .com, etc., individuals, etc.), and so on, by making requests of the server interactively, or by means of preferences registered with the server.

A refinement of the third approach is to build or generate a more helpful "key-phrase home page" on the link server. This page could provide for example, a brief definition, followed by an organized presentation of all available sources of information, organized by type and ranked according to quality in each case. In this format, the casual user may be satisfied with a mere dictionary-style definition, while the serious scholar can more readily pursue a wider variety of sources beyond merely the top-ranked source. Providing interactive reorganization of the definition page according to end user preferences is preferable in this case as well.

A fourth approach is to use JavaScript features to install a snapshot of the key-phrase home page at the time of link installation. In this approach, a JavaScript pop-up menu may hold a list of all competing links for the linked topic.

Example Key-Phrase Home Page Format

Below is an example of how a very simple "key-phrase home page" might be laid out:

                             TABLE 1
             Key Phrase: Dictionary-style definition
     Link to highest-rated online encyclopedia-style article
       Link to highest-rated online tutorial, if available
    Link to highest-rated textbook covering this topic, if any
    Link to educational resources (online courses, degree programs, etc.)
    Highest rated related links ("See also" type information)
          Rank-ordered list of encyclopedia-style links
              Rank-ordered list of online tutorials
          Rank-ordered list of other online information
      Rank-ordered list of contributed links of unknown type
                              . . .
          Last unrated contributed link of unknown type

        TABLE 2
        00    link all occurrence of a key phrase, whether emphasized
              or not;
        01    link all occurrences of a key phrase, but only when
              emphasized;
        10    link the first occurrence of a key phrase in each page
              (whether emphasized or not), and all emphasized
              occurrences; and
        11    link only the first emphasized occurrence of a key
              phrase.

    GROUPS = CM_DSP
    PATH = /Education/Technology
    KEY = W3K
    URL = http://www.w3k.org
    # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
     - - - - - - - - - - -
    KEY = / Legal / GNU General Public License
    URL = http://www.fsf.org/copyleft/gpl.html
    # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
     - - - - - - - - - - -
    PATH = /Music/Computer_Music/Signal_Processing/People
    KEY = Julius O. Smith III .vertline. Julius O. Smith .vertline. Julius
     Smith
    URL = http://www-ccrma.stanford.edu/.about.jos/
    KEY = JOS
    URL = http://www-ccrma.stanford.edu/.about.jos/
    # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
     - - - - - - - - - - -
    PATH = /Music/Computer_Music/Centers/CCRMA
    KEY = CCRMA Courses
    URL = http://www-ccrma.stanford.edu/CCRMA/Overview/courses.html
    KEY = CCRMA Research
    URL = http://www-ccrma.stanford.edu/CCRMA/Overview/research.html
    KEY = CCRMA Overview
    URL = http://www-ccrma.stanford.edu/CCRMA/Overview/Overview.html
    KEY = CCRMA
    URL = http://www-ccrma.stanford.edu/
    # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
     - - - - - - - - - - -
    PATH = /Music/Computer_Music/Sound_Synthesis
    SYNM = /Engineering/Signal_Processing/Sound_Synthesis
    KEY = Lagrange Interpolation
    URL = http://www-ccrma.stanford.edu/.about.jos/Lagrange_Interpolation.html
    LEVEL = 12
    KEY = Lagrange Interpolation
    URL = http://www.acoustics.hut.fi/.about.vpv/publications/vesa_phd.html
    KEY = Digital Waveguide Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/wg.html
    KEY = Commuted Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/book2000/
    CommutedSynth.html
    KEY = Virtual Analog Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/VirtualAnalog/
    VirtualAnalog.html
    KEY = Physical Modeling Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/pmupd/PMSynthesis.html
    # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
     - - - - - - - - - - -
    PATH = Music/Computer_Music/Signal_Processing/
    Sound_Synthesis/Vintage_Methods
    KEY = Additive Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/SMS_PVC/
    AdditiveSynth.html
    KEY = Sampling Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/samplingsynth.html
    KEY = Cross-Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/crosssynth.html
    # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
     - - - - - - - - - - -
    PATH=
    Music/Computer_Music/Signal_Processing/Sound_Synthesis/
    Acoustic_Instruments
    KEY = Bowed String Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/book2000/
    Bowed_Strings.html
    KEY = Brass Synthesis
    URL = http://www-ccrma.stanford.edu/.about.jos/pmupd/Brasses.html

        GROUPS = CM_DSP # Only operative if selecting based on group
        PATH = /Education/Technology
        PATH = /Music/Computer_Music/Signal processing/People
        PATH = /Music/Computer_Music/Centers/CCRMA
        PATH = /Music/Computer_Music/Sound_Synthesis
        SYNM = /Engineering/Signal_Processing/Sound_Synthesis
        PATH = Music/Computer_Music/Signal Processing/
        Sound_Synthesis/Vintage_Methods
        PATH=
        Music/Computer_Music/Signal_Processing/Sound_Synthesis/
        Acoustic_Instruments
    Adding some typical directives and eliminating one redundant
    specification leads to
        GROUPS = CM_DSP # Only operative if selecting based on group
        PATH = /Music/Computer_Music
        MULTIPLE_INHERITANCE_DEPTH = 2
        OWNERS_ONLY
        MIN_LEVEL = 12
        MAX_LEVEL = 100
        SUITABILITY = PG-13 # Movie and V-chip names understood
        SOURCE = ALL
        TYPE = Refereed
        PATH = /Education/Technology

    ID .vertline. KEY .vertline. URL .vertline. PropertyName:Value .vertline.
     PropertyName:Value .vertline. . . .
    ID .vertline. KEY .vertline. URL .vertline. PropertyName:Value .vertline.
     PropertyName:Value .vertline. . . .
    . . .

              23 .vertline. W3K .vertline. http://www.w3k.org .vertline.
     Date:2-Sep-99 .vertline.
                    Context : /Education/Technology/W3K .vertline.
                    Level:All .vertline. Rating:5 .vertline. RatingCount:7
     .vertline.
                    Hits: 20 .vertline. Installs: 4 .vertline.
                    Owner : Julius Smith .vertline.
                    Group : CM_DSP .vertline.
                    OwnerEmail : jos@w3k.org

    TABLE 3
    Property    Meaning
    Level       Educational level of the link, if not a range (1-100, All)
    MinLevel    Lower bound of educational level range, if applicable
    MaxLevel    Upper bound of educational level range, if applicable
    FullTitle   Contents of URL's HTML <TITLE> tag in quoted string
    Description Description of link by submitting user
    Date        Date link was submitted by user
    Type        Type of information (Encyclopedia, Tutorial, Book,
                Course, . . .
    Language    English, French, German, Spanish, . . .
    Suitability Similar to rating system used in the "V chip" for
                television
    Context     Context path (when handling many contexts per database
                file)
    Synonyms    List of equivalent phrases separated by `.vertline.`. Order is
                important.
    Hits        Number of times link accessed by browsing
    Installs    Number of documents link has been installed in
    Rating      Quality rating as a number from 1 to 10
    RateCount   Number of users contributing ratings
    isNew       1 if Date is sufficiently recent
    isPopular   1 if Hits is large relative to other links
    OwnerEmail  Email address of link contributor
    ReceiveMail 1 if link contributor wants our newsletter
    Groups      List of owning groups separated by `.vertline.`
    User1       Property defined by user
    User2       Property defined by user
    . . .       . . .