Homepage > Papers

Semantics representation

Kuznetsov Igor Petrovich. M. Nauka", 1986. 290 p.


Monograph consists of 7 chapters and applications. The purpose of monograph is the creation of united basis for the complex processing of different forms of knowledge, expressed in the forms of natural language (NL). Monograph develops the conception, which is consisted in the use for these purposes the uniform (gomogeneous) means, oriented to presentation and processing of structural information.

The first chapter is dedicated to the development of the principles of the system organization, which take into account the open nature of many subject areas and constant expansion of the decided tasks. For this aims the constant expansion of the functional possibilities of system are required. It is important that it must be achieved not by the labor-consuming programming, but by input of declarative information (in the convenient for the man forms) into system. Such information must be converted automatically into the internal knowledge representation. As a result the system must be tuned to one or other form of activity.

The second chapter is dedicated the selection of the language for knowledge representation with orientation for the autonomous inside system processing. The fact is that on the form of internal language construction depends the complexity of the working procedures, and respectively, of the realizing programs. With the complication of the language syntax is begun more difficultly to correct expressions, to identify semantic components and so on. Because the adequate means of knowledge presantation must be simple, uniform, i.e., satisfy the so-called criterion of uniformity. They must have structural basis - in order to reflect the structural nature of many subject areas. In more degree these criteria are satisfied the semantic networks. The problem of their further development to increase their depictive possibilities with presarvation of the important property of uniformity is solved.

For this aims we introduce and develop the language of the extended semantic networks (ESN). The procedures of their use for the presantation of different formalisms are considered, for example: set-theoretical model, the relational model of Data Base, constructions of algorithmic languages and predicate logic ets.

The third chapter is dedicated the creation of the standardized methods, which allow the complex decision of the tasks of knowledge inputing and using. Selection of internal language in many respects determines the procedures of working. For processing the gomogeneous representation (ESN) cannot be used the means of logical inference from the classical logic. Special methods are required. We introduce the new concepts for this: semantic graphs (ESN with the interpreting operators for search and refinement of information) and worknet productions (it is the rules IF THEN where the left and right parts are ESN).

In the fourth and fifth chapters we consider the use of the introduced graphs and productions for the solution of the problems of the intellectual systems: logical conclusion, the transformation of structure representations, active extraction of the missing and contradictory information. The great possibilities of the introduced means, which in many respects cover the existing formalisms, are illustrated.

In last chapters the formalisms are the subject of studies. Sign model is considered as the collection of symbols and actions above them, carried out by a mathematician. All this is represented in the formalism ESN and is investigated on the meta-level. It shows how formal records are converted to the semantic level (ESN) and as manipulation by them is achieved with the aid of the productions. On the sematics level are considered the standard paradoxes (Russell, liar, barber, linguistic paradoxes). Feedback and reasons for their appearance are studied. In the application linguistic processor on the network productions is proposed. Its work is reduced to the conversion of representations from the level of surface structures to the level of the semantic structures which are in the forms of ESN. This is new direction in the construction of hardware of intellectual systems.

Kuznetsov Igor Petrovich (igor-kuz@mtu-net.ru)