AUTOMATIC TEXT TRANSLATION SYSTEM FOR ARTIFICIAL LLANGUAGES

Abstract

The growing number and variety of artificial languages leads to the need and relevance of creating automatic dictionaries for their translation in order to facilitate human communication. Such languages include languages where vocabulary, phonetics, and grammar have been specifically designed to achieve specific goals and to communicate with a group of people by interests or place of residence. These languages can be distributed among people of certain professions or among neighboring nations. Examples are slang and surzhik. The common for them is that there is a basic language (literary), the intersection in spelling and meaning of words and phrases with which is quite large. The main goal of the project is to create a system of automatic translation of words and texts from / into arbitrary languages, including hybrid, artificial and slang ones. The proposed model shows the interaction and partial interdependence of the creation and adjustment modules and the translation module of the dictionary, which is explained by tacking the approach of reverse propagation of the translation error. To perform experiments and analyze the performance of the proposed approach to the organization of automatic translation of texts from and into arbitrary language, a software application was developed, which includes a subprogram of initial word processing for dictionary organization, one for creating a working dictionary and one for two-way improvement of created dictionary by the inclusion of new texts in order to improve the quality of translation, including the search for word phrases, idiom, and translation for them, the subprogram of dividing the dictionary into sub-dictionaries with a small percentage of text, the subprogram of the translator itself. To test and analyze the results of the proposed generalized model, three types of source texts were used: literary poetry translation, literary prose translation, literal prose translation. The results of the experiments showed that the proposed approach provides a high level of translation (up to 98,8%) in similar languages (between such languages as Ukrainian-Russian, or Ukrainian - Ukrainian-Russian surzhik wih equal word order in the sentence), especially with a literally translated source text. It has become known that the use of artistic texts to generate dictionaries is possible, but not very effective.