Semantics-based system for factual knowledge retrieval with

        Russian-English natural language interface

      

         Igor P. Kuznetsov,  Elena  B. Kozerenko,  Sharnin M. M.

   (Moscow, Institute for Informatics Problems of the Russian Academy of Sciences)

 

                                        RESUME

 

     An experimental system is considered employing common mechanisms for deep sentence analysis of the Russian and English natural language texts containing factual knowledge, with the subsequent mapping of linguistic and subject knowledge into unified knowledge base structures, i.e. semantic networks. The peculiarity of the system consists in wide use of word meanings in the course of analysis when their correspondence to the classes of the word meanings family tree is established. These meanings facilitate the decision of the ambiguity problems caused by polysemantic verbs, they allow to restore the information given on default.

     The system utilizes the similarity of lexical and syntactic constructions of the two languages which makes it possible to develop compact bilingual analysis algorithms on the base of the Russian language system release. The unified linguistic knowledge input script is used for both languages enquiring morphological, and semantic-syntactic characteristics of words as well as their meanings. For verbs the syntactic-semantic frames are required which are their government models with admissible word classes indicated. The input of word meanings in terms of another natural language is allowed, which made it possible to design the unified family tree of concepts constituting the kernel of the knowledge base.

     The system provides the response to factual knowledge queries stated in the free form in one of the two languages.