Extracting bilingual dictionary from comparable corpora with dependency heterogeneity

This paper proposes an approach for bilingual dictionary extraction from comparable corpora. The proposed approach is based on the obser-vation that a word and its translation share similar dependency relations. Experimental re-sults using 250 randomly selected translation pairs prove that the proposed approach signifi-cantly outperforms the traditional context-based approach that uses bag-of-words around translation candidates.