Constrained Hidden Markov Model for Bilingual Keyword Pairs Alignment

摘要

Bilingual terminology dictionaries are resources of much practical importance in many application of bilingual NLP. Because technical terminology can be both very specific and rapidly evolving, it can however be difficult to obtain dictionaries with good coverage. Mining automatically such terminology from technical documents is therefore an attractive possibility. With this goal in mind, and following some previous works, we devise an algorithm that is efficient at aligning the bilingual keyword list of scientific papers. Our results show that our approach can extract bilingual terms with very good precision and recall.