Thai Dependency Parsing with Character Embedding

Abstract

Dependency parsing (DP) becomes an important part of natural language processing (NLP) applications. However, most of DP methods have been developed for English language, but not for Thai language. In addition, the existing DP methods were still unsolved the problems of long and complex sentences. Therefore, this paper proposes seven Thai DP algorithms. Five different Thai DP algorithms was developed from transition-based parsing and the other two was developed from graph-based parsing. Based on Thai-PUD and English-PUD datasets, containing both long and complex sentences, the experimental results showed that all Thai DP algorithms bundled with character embedding can outperform the baselines.