Statistical Malay Dependency Parser for Knowledge Acquisition Based on Word Dependency Relation

Abstract

One of the common problems faced when processing information gathered from any natural language is the “semantic gap” where the “meaning” of the sentences is not exactly extracted. In Malay Natural Language Processing (NLP), as our knowledge, there is no existing Malay Parser that can be used to develop a knowledge acquisition feature to extract “meaning” from Malay articles based-on syntactic relations. This relation is basically the relation between a word and its dependents. This paper will examine the Dependency Grammar (DG) for developing Malay Grammar Parser and discuss the possibilities of developing probabilistic dependency Malay parser using the projected syntactic relation from annotated English corpus. The English side of a parallel corpus, project the analysis to the second language (Malay). Thus, the rules for adaptation from English DG to Malay DG will be defined. The projected tree structure in Malay will be used in training a stochastic analyzer. The training will produce a set of tree lattices which contains chunks of dependency trees for Malay attached with their probability value. A decoder will be developed to test the lattices. A DG for a new Malay sentence is built by combining the pre-determined lattices according to their plausible highest probability of combination.