Paper: Two Methods for Extending Hierarchical Rules from the Bilingual Chart Parsing

ACL ID C10-2021
Title Two Methods for Extending Hierarchical Rules from the Bilingual Chart Parsing
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010
Authors

This paper studies two methods for train- ing hierarchical MT rules independently of word alignments. Bilingual chart pars- ing and EM algorithm are used to train bi- text correspondences. The first method, rule arithmetic, constructs new rules as combinations of existing and reliable rules used in the bilingual chart, significantly improving the translation accuracy on the German-English and Farsi-English trans- lation task. The second method is pro- posed to construct additional rules directly from the chart using inside and outside probabilities to determine the span of the rule and its non-terminals. The paper also presents evidence that the rule arithmetic can recover from alignment errors, and that it can learn rules that are difficult to learn from bilingual alignments.