UNL-Based Bangla Machine Translation Framework

Abstract

The usage of native language through Internet is highly demanding due to the rapid increase of Internet-based applications in daily life. As information is available in the Internet in different languages, it is impossible to retrieve the information in other languages. Universal Networking Language (UNL) addresses this issue by converting the requested information from other languages to UNL expressions followed by UNL expressions to respective native languages. Even though Bangla is the sixth most popular language in the world, there is no system developed so far to convert Bangla text into UNL expressions and vice versa. For this purpose, the authors develop a framework. The framework has two constituent parts: 1) EnConverter: converts Bangla native sentences into UNL expressions considering UNL compatible Bangla word dictionary and analysis rules, and 2) DeConverter: converts UNL expressions into respective Bangla sentences considering Bangla word dictionary and generations rules. In both cases, case structure analysis, Bangla parts of speech, and different forms of verbs along with their prefixes, suffixes, and inflections are taken into consideration. This chapter describes the complete theoretical analyses of the EnConversion and DeConversion frameworks. The experimental results confirm that the proposed framework can successfully convert Bangla sentences into UNL expressions, and also can convert UNL expressions into corresponding Bangla sentences.