Application of Semantic Analysis to the Russian-to-English Machine Translation

This report focuses on the application of the professor Tuzov's theory to the design and elaboration of the Russian-to-English machine translation system.

Machine Translation(MT) becomes an integral part of our lives. The technologies of MT rose from the simplest light-weighted systems of direct translation using only morphological analysis to highly-developed powerful systems based not only on morphological but also on syntactic and even on semantic analysis. Despite intensive researches on this subject, existing systems of MT are still imperfect. And the problem does not consist in the translation of specific words or phrases, the problem is the inadequacy of semantic analysis algorithms. Meaning of the text still remains the mystery for computer. But how can the correct target-language word be selected without understanding semantics of the source text, considering that the majority of the high-usage words of any natural language are multivalued?

Owing to the works of professor V.A.Tuzov we have tools for Russian-language texts processing. Tuzov's analyser is capable of choosing correct semantic alternative for each word in sentence, and it can be successfully applied to solving the principal problems of Russian-to-English translation.

Quality translation requires some conversion dictionary, in which all semantic alternatives of any Russian-language word are putted in correspondence with their English-language equivalents. Construction of conversion dictionary forms special scientific problem. Key aspects of this problem are paid special attention in this report.

Suggested approach doesn't solve all problems of MT, but can contribute to overcoming of those difficulties which hinder Machine Translation to achieve brand new level.