Resolving Function Tagging Ambiguity in the Myanmar Language Using Transformation-Based Learning

Abstract

This article investigates the use of transformation-based learning for resolving function tagging ambiguity in the Myanmar language. Function tagger plays an important role in natural language applications like speech recognition, natural language parsing, information retrieval and information extraction. In this paper, the function tagger [12] learns rules to correct its mistakes. A set of rule templates is used to create specific rules. At initial stage of function tagging for Myanmar, it is trained with a very limited resource of annotated corpus. The performance can be maximized with a substantial amount of annotated corpus. The function tagset has been developed for training and testing the function tagger. The present tagset consists of 56 tags. A corpus size of about three thousand sentences is used for training and testing the accuracy of the function tagger. The tagger learned 192 rules (including lexical and contextual rules) and achieved 93% accuracy.