Extracting bilingual terminologies from comparable corpora

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 402–411, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics Extracting bilingual terminologies from comparable corpora Ahmet Aker, Monica Paramita, Robert Gaizauskas University of Sheffield ahmet.aker, m.paramita, r.gaizauskas@sheffield.ac.uk Abstract In this paper we present a method for extracting bilingual terminologies from comparable corpora. In our approach we treat bilingual term extrac- tion as a classification problem. For classification we use an SVM binary classifier and training data taken from the EUROVOC thesaurus. We test our approach on a held-out test set from EUROVOC and perform precision, recall and f-measure eval- uations for 20 European language pairs. The per- formance of our classifier reaches the 100% pre- cision level for many language pairs. We also perform manual evaluation on bilingual terms ex- tracted from English-German term-tagged compa- rable corpora. The results of this manual evalu- ation showed 60-83% of the term pairs generated are exact translations and over 90% exact or partial translations. 1 Introduction Bilingual terminologies are important for various applications of human language technologies, in- cluding cross-language information search and re- trieval, statistical machine translation (SMT) in narrow domains and computer-aided assistance to human translators. Automatic construction of bilingual terminology mappings has been investi- gated in many earlier studies and various methods have been applied to this task. These methods may be distinguished by whether they work on parallel or comparable corpora, by whether they assume monolingual term recognition in source and target languages (what Moore (2003) calls symmetrical approaches) or only in the source (asymmetric ap- proaches), and by the extent to which they rely on linguistic knowledge as opposed to simply statis- tical techniques. We focus on techniques for bilingual term ex- traction from comparable corpora – collections of source-target language document pairs that are not direct translations but are topically related. We choose to focus on comparable corpora because for many less widely spoken languages and for technical domains where new terminology is con- stantly being introduced, parallel corpora are sim- ply not available. Techniques that can exploit such corpora to deliver bilingual terminologies are of significant practical interest in these cases. The rest of the paper is structured as follows. In Section 2 we outline our method. In Section 3 we review related work on bilingual term ex- traction. Section 4 describes feature extraction for term pair classification. In Section 5 we present the data used in our evaluations and discuss our results. Section 6 concludes the paper. 2 Method The method we present below for bilingual term extraction is a symmetric approach, i.e. it assumes a method exists for monolingual term extraction in both source and target languages. We do not pre- scribe what a term must be. In particular we do not place any particular syntactic restrictions on what constitutes an allowable term, beyond the require- ment that terms must be contiguous sequences of words in both source and target languages. Our method works by first pairing each term …

Content truncated to 500 words