An unsupervised method for ranking translation words using a bilingual dictionary and WordNet

In the context of machine translation, picking the correct translation for a target word among multiple candidates is an important process. In this paper, we propose an unsupervised method for ranking translation word selection for Korean verbs relying on only a bilingual Korean-English dictionary and WordNet. We focus on deciding which translation of the verb target word is the most appropriate by using a measure of inter-word semantic relatedness through the five extended relations between possible translations pair of target verb and some indicative noun clues. In order to reduce the weight of application of possibly unwanted senses for the noun translation, we rank the weight of possible senses for each noun translation word in advance. The evaluation shows that our method outperforms the default baseline performance and previous works. Moreover, this approach provides an alternative to the supervised corpus based approaches that rely on a large corpus of senses annotated data.