OOV Term Translation, Context Information and Definition Extraction Based on OOV Term Type Prediction

Abstract

Although there are many existing approaches for solving the OOV term translation problems, but existing approaches are not able to handle different types of OOV terms, especially hybrid translations, such as “Kenny-Caffey syndrome (Kenny-Caffey 氏症候群)”. We proposed a novel integrated ranking approach to consider the types of OOV terms before translating them. Thus, different types of OOV terms could be translated differently. Furthermore, the translations mined in other languages are also OOV terms, none of existing approaches offer the context information or definitions of the OOV terms. Users without special knowledge cannot easily understand meanings of the OOV terms. Our integrated ranking approach also extracts monolingual definitions and multilingual context information of OOV terms. Moreover, we propose a novel adaptive rules approach with Bayesian net and Adaboost for handling hybrid translations. Experiments show our approach performs better than existing approaches.