An Automatic Bilingual Corpora Generator

Abstract

Bilingual corpora that contains similar documents of two different languages are examples of essential resources for Natural Language Processing (NLP) tasks including Cross-Lingual Information Retrieval (CLIR) and machine translation. Nevertheless, these resources could also be useful for many processes in learning languages. We introduce an automatic bilingual corpora generator that builds corpus resources from the web. This generator involves the use of the in-domain terms (IDT), in which the terms can be thought of as the most important contextually relevant words. The method used is simple yet practical, and makes acquiring resources from web sources more than just collecting texts and pasting them all together. However, as an on-going project, the system has not been fully implemented and evaluated. In this paper, the researchers emphasizes more on the prototype of the system in terms of appearance and display. For example, the generator shall be built on a web-based system that gives different options to users on how they would like to observe the acquired texts.

Keywords

• bilingual corpora
• in-domain-term (idt)