L&H Lexicography Toolkit for Machine Translation

Abstract

One of the most important components of any machine translation system is the translation lexicon. The size and quality of the lexicon, as well as the coverage of the lexicon for a particular use, greatly influence the applicability of machine translation for a user. The high cost of lexicon development limits the extent to which even mature machine translation vendors can expand and specialize their lexicons, and frequently prevents users from building extensive lexicons at all. To address the high cost of lexicography for machine translation, L&H is building a Lexicography Toolkit that includes tools that can significantly improve the process of creating custom lexicons. The toolkit is based on the concept of using automatic methods of data acquisition, using text corpora, to generate lexicon entries. Of course, lexicon entries must be accurate, so the work of the toolkit must be checked by human experts at several stages. However, this checking mostly consists of removing erroneous results, rather than adding data and entire entries. This article will explore how the Lexicography Toolkit would be used to create a lexicon that is specific to the user’s domain.