Back-Propagation Neural Network Approach to Myanmar Part-of-Speech Tagging

摘要

Part-of-Speech (POS) tagging is the process of assigning a POS label to each of a sequence of words. It is also a lowest level of syntactic analysis and useful for many natural language processing (NLP) tasks such as subsequent syntactic parsing and word sense disambiguation. We developed an annotated corpus and POS tagger for Myanmar language based on back-propagation neural network (BPNN) model. In our experiments, BPNN model is trained with 3gram, 4gram and 5gram. The results show that the BPNN model with 4 g is able to achieve considerable higher F-scores on the POS tagging task than 3 g and 5 g models for both close and open test sets. Moreover, BPNN POS tagging approach performed better than proposed HMM with rule based.