Learning the Optimal Use of Dependency-parsing Information for Finding Translations with Comparable Corpora
Daniel Andrade, Takuya Matsuzaki, Jun’ichi Tsujii · 2011
Using comparable corpora to find new word translations is a promising approach for ex-tending bilingual dictionaries (semi-) auto-matically. The basic idea is based on the assumption that similar words have similar contexts across languages. The context of a word is often summarized by using the bag-of-words in the sentence, or by using the words which are in a certain dependency position, e.g. the predecessors and succes-sors. These different context positions are then combined into one context vector and compared across languages. However, previ-ous research makes the (implicit) assumption that these different context positions should be weighted as equally important. Furthermore, only the same context positions are compared with each other, for example the successor po-sition in Spanish is compared with the suc-cessor position in English. However, this is not necessarily always appropriate for lan-guages like Japanese and English. To over-come these limitations, we suggest to perform a linear transformation of the context vec-tors, which is defined by a matrix. We de-fine the optimal transformation matrix by us-ing a Bayesian probabilistic model, and show that it is feasible to find an approximate solu-tion using Markov chain Monte Carlo meth-ods. Our experiments demonstrate that our proposed method constantly improves transla-tion accuracy. 1