Problematics of the Systems Understanding Natural Languages

Abstract

Artificial intelligence (AI) researchers are concerned with developing computer systems that simulate human intelligence using techniques of semantic processing. AI is divided into three independent research areas: Natural Language Processing (NLP), robotics and expert systems. The communication is dedicated to the systems understanding natural languages. The problems necessary for the elaboration of such systems are: the domain of knowledge, the process of knowledge representation, organization of control of actions, organization of knowledge and access to it. If taking each of the problems apart we can say that the selection of the types of presentations of knowledge is unique but the means of presentations are different. Because of a big volume of informational data it is necessary to introduce an interpretation regime and an archives service. In such a case the system must possess metaknowledge of three types: metaknowledge of the environment, metaknowledge about the language of communication and metaknowledge about the participants of communication. The management of the intercourse system is performed by such methods as „Top-down” and „Bottom-up”. As it is impossible to prepare all the necessary reactions of the system, the only way out consists in the fact that the system itself realizes what happens and is able to express it in terms understood by the user. Such selfconsciousness of the system sometimes is very difficult to organize because of a big volume of data. It all depends on the access of knowledge which must be given in alphabetic order, small units must be transformed into bigger ones called blocks. New processed information is received by concluding rules and facts stored in the system by means of formal, specialized and metaconclusions. These bring to a good system understanding natural languages which then must be translated by means of the computer programs.