Paper: On-line Language Model Biasing for Statistical Machine Translation

ACL ID P11-2078
Title On-line Language Model Biasing for Statistical Machine Translation
Venue Annual Meeting of the Association of Computational Linguistics
Session Main Conference
Year 2011
Authors

The language model (LM) is a critical com- ponent in most statistical machine translation (SMT) systems, serving to establish a proba- bility distribution over the hypothesis space. Most SMT systems use a static LM, inde- pendent of the source language input. While previous work has shown that adapting LMs based on the input improves SMT perfor- mance, none of the techniques has thus far been shown to be feasible for on-line sys- tems. In this paper, we develop a novel mea- sure of cross-lingual similarity for biasing the LM based on the test input. We also illustrate an efficient on-line implementation that sup- ports integration with on-line SMT systems by transferring much of the computational load off-line. Our approach yields significant re- ductions in target perplexity compared to t...