Method and system for processing, storing, retrieving and presenting information with an extendable interface for natural and artificial languages6999934Abstract A method and system for processing, storing, retrieving and presenting information with an extendable interface for natural and artificial languages. The system includes an interpreter, a knowledge base, and an input/output module. Making use of an internal representation based on sorted-type theory, the system stores information in the knowledge base, answers queries from clients, and processes erroneous or contradictory information according to a dynamically adjustable set of rules. The system also stores language definitions in the knowledge base, enabling the system to communicate with clients in a variety of natural and artificial languages. New languages may be added to the system by presenting definitions expressed in a language already incorporated within the system. Claims What is claimed is: Description FIELD AND BACKGROUND OF THE INVENTION
These features are useful in applications in various industrial, commercial, social, scientific and educational domains, especially those characterized by:
The present invention makes use of Sorted-Type Theory. See for example, Daniel Gallin, Intensional and Higher-Order Modal Logic, North Holland Publishing Company (an imprint of American Elsevier Publishing Company), Amsterdam, 1975, ASIN: 044411002X, which is incorporated by reference for all purposes as if fully set forth herein. See also, for example, B. Carpenter, Type-logical Semantics, MIT Press, Cambridge, Mass., 1997, ISBN: 0262531496, which is incorporated by reference for all purposes as if fully set forth herein. A brief introduction to sorted-type theory follows: A.1) Tyn language: Basic concepts of any type theory are types and terms. In the n-sorted-type theory there are n+1 primitive types:
For example, using the synonyms e,f, and g for e1, e2,, and e3,, respectively, (et), t(ef), ((et)g) and (((et))(tg)) are all derived types of Ty3. For the sake of brevity, we will omit below parentheses in functional types unless they are required to express pairing in an order other than from right to left, i.e., the types of this example would be written as: et, tef, (et)g, and ((et)f)tg. Terms of Tyn are characterized recursively: primitive terms of type a are variables xa0, xa1, xa2, . . . and constants Ca0, Ca1, Ca2 . . . (the first three variables of type a will be also denoted by synonymous symbols xa, ya, za); derived terms are built by means of the three fundamental operations of Tyn: if Y is a term of type ab and X is a term of type a, then application Y X is a term of type b; if Y is a term of type b, then lambda abstraction λxakY is a term of type ab; if X and Y are terms of type a, then equality X=Y is a term of type t. Examples of Ty3 terms include: xe1, C(et)f12, C(et)f12xet4, λxet4cf6, xef2=Cef0, and Cte5(Cgf1=λxgCf9). Below, important notions of free and bound occurrence of a variable in a term are used: an occurrence of a variable xb in a term Aa is bound if it occurs only within a part like λxbBc, otherwise it is said to be free. All simpler terms incoming as constituents into a derived term (except bound variables) are said to be its sub-terms. For example, sub-terms of an application YX or an equality X=Y are X, Y and all sub-terms of X and Y (recursively); sub-terms of λxakY are Y and all its sub-terms. We will denote the fact that a term B is a sub-term of A with the notation A(B). Given a term A(xa) and a term B, we say that B is free for xa in A(xa) if no free occurrence of xa in A(xa)lies within a part λybC, where yb occurs free in B. In other words, B is free for xa in A(xa) if and only if no one of free variables of B proves to be bound when B replaces all free occurrences of xa in A. A.2) Tyn Semantics: Exploring now the meaning of Tyn terms, consider n arbitrary (but non-empty) sets M1, M2, . . . Mn and also a special set 2={0, 1} (note the underlining which serves to distinguish this notation from the numeral 2). Let us associate with any type a some set Ma defined recursively as follows: Mt=2, Me=M1, Mf=M2, and so on,
Let any constant of type a represent a certain element of Ma, and any variable of type a represent an arbitrary ("unknown") element of Ma. Then we can (again recursively) specify what is represented by any term: Y X represents the result of the application of function [Y] to argument [X], where [Aa] denotes an element of Ma represented by a term; λxakY represents the function from Mab whose value for any X from Ma is equal to [Y](xak/X), where [Y](xak/X ), denotes that element of Mb which is represented by Y, subject to the condition that [xak]=X; X=Y represents 1 if [X]=[Y] and 0 otherwise. Thus, having assigned each Tyn constant Cak an element [Cak] of a corresponding set Ma, we can find an element [A] of a corresponding set for any given term A and for any given assignment to the Tyn variables. A mapping M: Cak→[Cak] (for given sets M1, M2, . . . Mn) is said to be a model of Tyn based on M1, M2, . . . Mn. Now, a few new, simple but important definitions:
Informally, the above construction can be interpreted as follows: ei-terms (that is, terms of type ei) represent entities of the i-th sort (either constant or depending on some variable objects) and t-terms represent truth values (also either constant or depending on some variable objects); ab-terms generally represent functions from a-objects to b-objects, for example, at-terms represent predicates about a-objects. It is easy to check that the formulas: T≡(λxtxt=λxtxt) F≡(λxtxt=λxtT) (where the sign "≡" serves to introduce synonymous notations) in any model are represented by 1 and 0, correspondingly; that is why they define regular "true" and "false" sentences of the first-order logic. Similarly the regular sentential connectives and quantifiers can be defined:
The power of Tyn (as well as of any other formal theory) is revealed by the fact that it is sufficient to supply a very short list of some special formulas (referred to as axioms) and inference rules in order to be able to obtain potentially infinite set of formulas which will be true in any model. Following is the list of axioms of Tyn: fttT ##CHR4## fttF=Axt(fttxt) 1. xa=ya→fatxa=fatya 2. Axa(fabxa=gabxa)=(fab=gab) 3. (λxaAb(xa))Ba=Ab(Ba) 4. where Ab(Ba) comes from Ab(xa) by replacing all free occurrences of xa by the term Ba, and Ba is free for xa in A(xa). Tyn has a single inference rule:
A proof in Tyn is a sequence of formulas each of which is either an axiom or else is obtained from earlier formulas by the inference rule. A formula A is said to be provable or a theorem of Tyn, which fact is denoted as |-A, if there is a proof in which A is the last formula. A formula A is said to be derivable from a set G of formulas if it is provable in Tyn supplied with all formulas from G as additional axioms, which fact is denoted as G |-A. A set G of formulas is said to be consistent if F is not derivable from G. Finally, if |-˜A, then A is said to be refutable; we also say that A is refutable by G if G |-˜A. It is easy to prove that, if A is refutable by G, then G ##CHR5## {A} is inconsistent. Now we are ready to formulate the two fundamental facts:
A final note: it is not to be supposed that any particular formula is either provable or refutable in Tyn: in fact there is an infinite number of formulas that are neither provable nor refutable. This important fact allows extensions to the set of the common Tyn axioms by infinitely large consistent sets of additional, specific axioms. A theory obtained as the result of such an expansion of the axiom set is said to be a restriction of Tyn because the sets of provable and refutable formulas for the new theory, of course, contain more formulas. Various attempts have been made to provide systems that process information in a variety of natural languages. U.S. Pat. No. 6,182,062 presents a system that coexistently stores information in a variety of languages. In this approach, the amount of storage needed increases in proportion to the number of languages. There is thus a widely recognized need for, and it would be highly advantageous to have, an information processing system able to communicate in a variety of natural and artificial languages, adopt new languages, store information in a single internal representation, and respond to requests for information in a language chosen by the client. SUMMARY OF THE INVENTION According to the present invention there is provided a method for information processing including the steps of: (a) providing an interpreter, the interpreter including: (i) a driver, and (ii) a language processor; (b) providing a knowledge base operative to store an internal representation of information; (c) providing a knowledge base manager; (d) providing an input/output module; (e) the interpreter converting, using the language processor, input information presented to the input/output module, and expressed in an external representation, into the internal representation; (f) the driver transmitting the internal representation of the input information to the knowledge base manager, and (g) the knowledge base manager modifying the knowledge base according to the knowledge base and according to the input information. Preferably, in the method, the information stored in the internal representation in the knowledge base includes a definition of a derived language. Preferably, in the method, the information stored in the internal representation in the knowledge base includes target information. Preferably, in the method, the method further includes the step of: (h) converting the input information into the internal representation according to the definition of the derived language. Preferably, in the method, the method further includes the step of: (h) using the definition of the derived language to convert information expressed in the internal representation into the derived language. Preferably, in the method, the method further includes the step of: (h) using the definition of the derived language to convert a definition of an additional derived language, presented to the input/output module in the derived language, into a definition, in the internal representation, of the additional derived language. Preferably, in the method, the internal representation includes entities logically equivalent to terms of a sorted-type theory. Preferably, in the method, a set of primitive types of the sorted-type theory includes a symbolic type, and a constant of the symbolic type represents an elementary symbol of a language, and a function involving the symbolic type is operative to represent an aggregation of the elementary symbols. Preferably, in the method, the set of primitive types of the sorted-type theory includes a meta-context type, and a constant of the meta-context type represents a meta-context including a set of non-logical axioms of the sorted-type theory, the meta-context further including a setting operative to control a behavior of a system, and a constant of functional type involving the meta-context type represents an operation on the meta-context. Preferably, in the method, the method further includes the step of: (h) organizing the information in the internal representation stored in the knowledge base as a set of the meta-contexts. Preferably, in the method, the method further includes the step of: (h) converting the input information into a sequence of operations on the meta-contexts. Preferably, in the method, the operation includes an operation selected from the group consisting of adding an axiom to a meta-context, removing an axiom from a meta-context, modifying an axiom in a meta-context, retrieving an axiom from a meta-context, outputting information in an external representation via the I/O module, and changing the settings in a meta-context. Preferably, in the method, the knowledge base includes a definition of a derived language, and the method further includes the step of: (h) if the input information includes, in a context defined in the definition of the derived language, a phrase undefined in the definition of the derived language, translating the phrase, according to the context, to at least one non-primitive term of the sorted-type theory, the term containing the phrase as at least one sub-term of a symbolic type. Preferably, in the method, the method further includes the steps of: (i) if, in a new context, a phrase previously stored in the knowledge base as at least one non-primitive term is found, generating a new axiom of the sorted-type theory, and j) causing the knowledge base manager to replace the at least one non-primitive term with other terms. Preferably, in the method, the knowledge base includes a definition of a derived language, and the method further includes the steps of: (h) if the input information includes, in a context defined in the definition of the derived language, a phrase undefined in the definition of the derived language, generating a new axiom of the sorted-type theory, and (i) adding the new axiom to the knowledge base. Preferably, in the method, the knowledge base includes a definition of a derived language, and the method further includes the steps of: (h) if the input information includes, lo in a context defined in the definition of the derived language, a phrase undefined in the definition of the derived language, requesting clarifying input from a source selected from the group consisting of a client and a server, and (i) modifying the knowledge base according to the knowledge base and the clarifying input. Preferably, in the method, the method further includes the steps of: (h) if the input information includes a query, the knowledge base manager processing the query according to the knowledge base and according to the query, and (i) transmitting to the driver, in the internal representation, a response to the query. Preferably, in the method, the method further includes the step of: (h) if the conversion of the input information into the internal representation causes an error condition, storing a portion of the input information in the knowledge base as at least one term of a symbolic type. Preferably, in the method, the method further includes the steps of: (h) detecting a contradiction between the input information and the knowledge base, and (i) modifying the knowledge base according to a set of error-handling rules stored in the knowledge base. According to the present invention there is provided an information processing system including: (a) an interpreter, the interpreter including: (i) a driver, and (ii) a language processor; (b) a knowledge base operative to store an internal representation of information; (c) a knowledge base manager, and (d) an input/output module, the system operative to convert, using the language processor, input information presented to the input/output module, and expressed in an external representation, into the internal representation, the driver operative to transmit the internal representation of the input information to the knowledge base manager, the knowledge base manager being operative to modify the knowledge base according to the knowledge base and according to the input information. According to the present invention there is provided a machine readable storage medium having stored thereon machine executable instructions, the execution of the machine executable instructions implementing a method for information processing, the method including the steps of: (a) providing an interpreter, the interpreter including: (i) a driver, and (ii) a language processor; (b) providing a knowledge base operative to store an internal representation of information; (c) providing a knowledge base manager; (d) providing an input/output module; (e) the interpreter converting, using the language processor, input information presented to the input/output module, and expressed in an external representation, into the internal representation; (f) the driver transmitting the internal representation of the input information to the knowledge base manager, and (g) the knowledge base manager modifying the knowledge base according to the knowledge base and according to the input information. The system of the present invention exchanges with users and/or with other external agents (systems, programs, devices etc.) various kinds of information:
The method of this invention is based on internal conversion of all the above kinds of information to sets of terms of a sorted-type theory [Gallin, Carpenter] with a set of primitive types including one or more special symbolic types, whose constants represent elementary symbols (such as letters or graphemes) of the system interface languages. For example, constant C0s, where s denotes a symbolic type, may stand for the letter 'A', C1s—for 'B' and so on. In examples below these constants are denoted by the corresponding letters in quotes. Constants of functional types involving only the symbolic type(s) represent various operations, such as concatenation, attachment, insertion, etc., to build strings, images or any other aggregations of the elementary symbols. For example, if the constant Csss stands for the operation of concatenation from left to right, then the term Csss 'm' (Csss 'a' 'n') represents the string "man" (in our notation of functional types, primitives are assumed to be associated in pairs from right to left, so that the type sss actually stands for (s(ss)) in the traditional notation of Gallin). In further examples such concatenated strings are simply denoted by double quoted strings like "man". Constants of functional types combining symbolic and non-symbolic primitive types represent relationships of arbitrary order between the symbolic terms and terms of other types. For example, constant Nouns(et)t may stand for the relationship between English nouns and their denominators, whose elements are to be introduced by axioms such as: Nouns(et)t"man" C1et where e and t denote individual and truth primitive types correspondingly. Similarly an axiom involving higher-order relationships may introduce, for example, the top-level structure of an English affirmative statement: Affirmstt=R(se(et)t)(e(et)t)stt(J(set)(s(et)t)se(et)t NPset VPs(et)t)λxeλxet(xetxe) where constants R(sabt) (abc)sct and J(sat)(sbt)sabt for any types a, b, c are defined to satisfy axioms: J(sat)(sbt)sabt psat qxbt(Csssxsys)xayb=(psatxsxa (R(sabt) (abc)sct psabt rabc)xs(rabcxaxb)=psabt xsxaxb and constants NPset and VPs(et)t are assumed to define noun and verb phrases of the statement, correspondingly. Hierarchical sets of such axioms may form definitions of various artificial and natural languages, referred to herein as derived languages. Some derived languages may be defined and used especially in order to specify definitions of other derived languages. In order to initiate this process, that is to specify the very first derived language, the method of this invention also assumes employment of a built-in, i.e. a pre-defined language which must be capable of expressing an arbitrary term of the sorted-type theory. For example, the standard type theory notation [Gallin], or its adaptation used in the present description, might serve as such a built-in language. The set of primitive types of the sorted-type theory may also include one or more special meta types allowing expression of some operations on sets of terms and/or other data and/or physical resources of the system. For example, a meta type m may stand for a pair of a set of axioms and an associated set of external devices, which pair is briefly referred to below as context. Terms of type mm then express various instructions that may be exchanged with the system as requests and responses—both can be directed from users and/or external agents to the system as well as in the opposite direction. For example, constants Asserttmm and Refutetmm may form instructions to add and delete an axiom to/from a given context. Another constant—Testtmm might form an instruction to test an axiom against a context, i.e. to attempt to infer (prove) or refute it or some instance of it from the axiom set of this context and output to a certain external device a response depending on results of this attempt. Other instructions may control the overall system behavior, for example, ReadIn(s(mm)t)mm might instruct the system to switch to use of another input language. Instructions may also be independent of the axiom set of a given context, for example, Sendsmm might instruct the system to output a given term in some external representation to a certain external device, and constant Ejectmm to simply execute a certain physical action. Combining meta and symbolic types enables implementation of context-dependent languages. Statements of such a language, as well as some of their members, should be translated to some combined instructions, while some other constituents, such as pronouns, should be converted to functions from m to other types. By means of the logical constant Paata defined to satisfy the axioms PaataxayaF=xa, PaataxayaT=ya, where F and T stand for the false and true constants, respectively, and the operation of the superposition function, Super(ab)(bc)ac=λyab λxbc λxa(xbc(yabxa)), arbitrarily complex instructions, i.e., programs, may be formed from a set of elementary (primitive) instructions. A major innovative feature of the present invention is the employment of an internal representation of information based on a specialized version of sorted-type theory with a set of primitive types including at least one symbolic type whose constants represent elementary symbols, such as letters or graphemes, of the system interface languages (for example, all Unicode codes). This crucial innovation implies the following unique and useful features that might not be achieved otherwise:
BRIEF DESCRIPTION OF THE DRAWINGS The invention is herein described, by way of example only, with reference to the accompanying drawings, wherein: FIG. 1 is a schematic illustration of an information processing system according to the present invention; FIG. 2 is a schematic illustration of a storage medium containing instructions to implement the present invention. DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention is of a method and system for processing information which can be used by clients to store, process, and retrieve information without having to know a particular database language or natural language. Specifically, the present invention can be used to gather information in a variety of natural and artificial languages, store the information in an internal representation, and, in response to queries about that information, presented in any of a variety of languages, provide answers in a language chosen by the client. The principles and operation of an information processing system according to the present invention may be better understood with reference to the drawings and the accompanying description. Referring now to the drawings, FIG. 1 illustrates schematically a preferred embodiment of an information processing system according to the present invention. Major components of the system include:
The modules of the system perform the following functions.
The system as a whole functions in the following way. Upon starting up, at least one input channel 42 is assigned to input information expressed in a particular built-in or previously-defined derived language and at least one output channel 44 is assigned to output system responses expressed in a built-in or previously-defined derived language. The language of an output channel 44 assigned to a particular dialog may or may not be the same as the language of an input channel 42 assigned that same dialog. The system starts to read input information, translate it to internal instructions, and execute these internal instructions. Execution of an instruction generally results in one or more of the following:
In particular, a sequence of instructions may create in the system Knowledge Base 10 a definition of a new language which then may be assigned to an input channel 42 or an 10 output channel 44. Lexemes and syntactic constructs of a derived language may serve as shortcuts for arbitrarily complex programs, i.e. combinations of instructions. Storing new primary content or meta information in the system is reduced to adding a new set of axioms to Knowledge Base 10. Contradictions and inconsistencies, as well as any other errors encountered during this process, are handled in accordance with currently set rules. Encountering in the input information from an input channel 42 fragments which cannot be parsed in the particular language assigned to that input channel 42 does not prevent complete storage of this information in the system because such fragments may be still stored in Knowledge Base 10 in the form of terms of a symbolic type. Moreover, if later the language definition in the system is extended, these fragments may be re-parsed and converted to new axioms in Knowledge Base 10. Other information processing operations, such as searching, merging, comparison, classification, refinement, translation, etc., are performed by programs that might be associated with certain queries in specially-defined input languages or entered by a user or another agent as a sequence of instructions or/and queries. Among external clients or agents with which the system exchanges information there might be other instances of the same system. This allows sharing and merging knowledge and information that might relate to different application areas, the information being accumulated independently and, possibly, initially expressed in different terminology or even different languages. An information processing system according to the present invention may be implemented as illustrated schematically, by way of example only, in FIG. 2. A computer 54 executes machine executable instructions 52 stored in machine readable storage medium 50. Machine readable instructions 52 are selected, in accordance with that which is taught in the present invention, such that execution of machine readable instructions 52 by computer 54 is operative to process information as described above. Machine readable storage medium 50 can include, but is not limited to, a memory device, such as a ROM or RAM, an optical storage medium, such as a CD-ROM or DVD, or a magnetic storage medium, such as a disk or tape. Optionally, machine readable storage medium 50 can be accessed remotely via a communication link. Many alterations and modifications of the information processing system illustrated in FIG. 2 may be made within the scope of the present invention. It is to be understood that the example of FIG. 2 is presented herein by way of illustration only, and is in no way intended to be considered limiting. While the invention has been described with respect to a limited number of embodiments, it will be appreciated that many variations, modifications and other applications of the invention may be made.
|
Same subclass Same class Consider this |
||||||||||