Syntactical Parsing for Ayta Abellen using PAWS generated Phrase Structure Rules

Automated syntactic parsing of Philippine languages could be foundational to future machine translation systems. Rule based systems for Philippine languages have typically not reached a level of wide coverage of language phenomena. The syntax parsing system described here uses the PAWS (Parser and Writer of Syntax) expert system to generate phrase structure rules. After customized rules common to most Philippine languages were added in the process of bringing a training data set up to 100% parsing, the auto-generated phrase structure rules were able to produce a correct parse for 81% of sentences in two Ayta Abellen native authored running texts. The customizations made for the training set helped further the development of the PAWS expert system for use with Philippine languages. The 81% parsing rate is significant in that it represents a wide range of coverage for a rules-based system.
5 pages
natural language processing
Subject Languages:
Content Language:
Nature of Work: