“VitaminE” System for Improving the Quality of Machine Translation.

Everyone who has experience of communication with Promt, Stylus, Socrat, and the like systems of machine translation are fully aware of the low quality of automatically translated texts. The low quality of translation stems from the fact that the contemporary systems for automatic translation of texts in natural languages process texts at the morphological and syntactic analysis levels while they are unable to work at the semantic level.  Separated from the context such translation yields a nonsense like “Часові мухи люблять стрілу” for the source English phrase “Time flies like arrow” instead of “Час летить як стріла”. Without adequate processing of the semantic (meaningful) level of a text, quality text translation is a far cry from the reality.

Where they fail to define unequivocally adequate translation, modern machine translations suggest a number of alternatives for the source word translation. “VitaminE” system which carries out semantic analysis of the meaningful context chooses out of possible alternative version of a correct one, which significantly improves the quality of machine translation.

While “VitaminE” system was being created, a number of efficient algorithms for bilingual associative and semantic analysis were developed to find a degree of semantic proximity between words as semantic objects.

When processing a text, the semantic analysis procedures interact with the developed bilingual semantic data base of knowledge  RusWordNet.

The semantic algorithms deal with semantic ambiguities of translation and cope with the problem of choosing the right alternative for the target text.