Paper: Machine Transliteration: Leveraging on Third Languages

ACL ID C10-2165
Title Machine Transliteration: Leveraging on Third Languages
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010
Authors

This paper presents two pivot strategies for statistical machine transliteration, namely system-based pivot strategy and model-based pivot strategy. Given two independent source-pivot and pi- vot-target name pair corpora, the mod- el-based strategy learns a direct source- target transliteration model while the system-based strategy learns a source- pivot model and a pivot-target model, respectively. Experimental results on benchmark data show that the system- based pivot strategy is effective in re- ducing the high resource requirement of training corpus for low-density lan- guage pairs while the model-based pi- vot strategy performs worse than the system-based one.