2009/016 - Source-Language Entailment Modeling for Translating Unknown Terms
Lucia Specia, Shachar Mirkin, Ido Dagan, Idan Szpekti, Marc Dymetman, Nicola Cancedda
ACL-IJCNLP 2009 (Association for Computational Linguistics and International Joint Conference on Natural Language Processing), Suntec, Singapore, 2-7 August 2009
We address the task of handling unknown terms in SMT, suggesting a more extensive use of source-language monolingual resources. We present a conceptual extension to prior work by allowing translations of entailed texts rather than exact paraphrases only. We further suggest a method for performing this process efficiently. Our experiments show that the proposed approach substantially increases translation coverage while maintaining translation quality.