A new algorithm and other software for disambiguation of polysemy and homonymy for computer translation into Russian Sign Language based on a semantic principle

Abstract

The paper analyses current computer Sign Language translation systems. Their advantages and disadvantages are detected. The main drawback is the lack of original text semantic analysis module capable of solving the task of disambiguation. A general scheme of translation system from phonic Russian to Russian Sign language including a module for semantic analysis is presented. It includes a block of source code analysis, developed by the authors, responsible for handling the semantic component of the Russian language. The semantic module relies on Tuzov’s dictionary. The semantic analysis algorithm is also described. The text analysis is completed when each word gets only one semantic description thus solving the problem of ambiguity. The most important developments of the semantic analysis module include the following: expanded collection of gestures, parsing of complex sentences, account in the algorithm analyses predicates classifier of Russian Sign Language. Testing of algorithm is made. The article compares the existing systems of computer translation from phonic to the sign language. The advantages and disadvantages of the considered systems are revealed and a conclusion is made about the need to take into account the semantic aspect of the translation process. A technology of semantic analysis is suggested. The model to choose an adequate meaning of a polysemic word or homonym on the basis of the automatic text processing system «Dialing» is described. Examples of the use of the software are given. The questions of testing the working capacity of the semantic analysis module are given due attention too. To enhance its efficiency, the system of semantic analysis was added to the translation system «Surdophone». To verify the efficiency of the semantic module’s operation, a comparison is made with the definition of some words’ semantic meanings by the systems «Yandex Translator» and «Google Translator». The present system showed its advantage in more complex cases. Also, the base of gestures of the RSL whose names are homonyms and polysemic words of the Russian language, were added and the features of their performance were revealed.