A Machine-Aided Approach to Generating Grammar Rules from Japanese Source Text for Use in Hybrid and Rule-based Machine Translation Systems

Author

Sean Michael Jones, University of Wisconsin-MilwaukeeFollow

Date of Award

December 2015

Degree Type

Thesis

Degree Name

Master of Science

Department

Computer Science

First Advisor

Susan McRoy

Committee Members

Ichiro Suzuki, Nicholas Fleisher

Keywords

Computational Linguistics, Machine Translation, Rule Based Machine Translation

Abstract

Many automatic machine translation systems available today use a hybrid of pure statistical translation and rule-based grammatical translations. This is largely due to the shortcomings of each individual approach, requiring a large amount of time for linguistics experts to hand-code grammar rules for a rule-based system and requiring large amounts of source text to generate accurate statistical models. By automating a portion of the rule generation process, the creation of grammar rules could be made to be faster, more efficient and less costly. By doing statistical analysis on a bilingual corpus, common grammar rules can be inferred and exported to a hybrid system. The resulting rules then provide a base grammar for the system. This helps to reduce the time needed for experts to hand-code grammar rules and make a hybrid system more effective.

Recommended Citation

Jones, Sean Michael, "A Machine-Aided Approach to Generating Grammar Rules from Japanese Source Text for Use in Hybrid and Rule-based Machine Translation Systems" (2015). Theses and Dissertations. 1058.
https://dc.uwm.edu/etd/1058