Semi-Automatic Identification of Bilingual Synonymous Technical Terms from Phrase Tables and Parallel Patent Sentences

In the research field of machine translation of patent documents, the issue of acquiring technical term translation equivalent pairs automatically from parallel patent documents is one of those most important. We take an approach of utilizing the phrase table of a state-of-the-art phrase-based statistical machine translation model. In this task, we consider situations where a technical term is observed in many parallel patent sentences and is translated into many translation equivalents. We apply SVM to the task of identifying synonymous translation equivalent pairs and achieve almost 98% precision and over 40% Fmeasure. Then, in order to improve recall, we introduce a semi-automatic framework, where we employ the strategy of selecting more than one seeds for each set of candidates bilingual synonymous term pairs. By manually judging whether each pair of two seeds is synonymous or not, we achieve over 95% precision and 50% recall.