Paper: Combining Context Features by Canonical Belief Network for Chinese Part-Of-Speech Tagging

ACL ID I08-2130
Title Combining Context Features by Canonical Belief Network for Chinese Part-Of-Speech Tagging
Venue International Joint Conference on Natural Language Processing
Session Main Conference
Year 2008
Authors

Part-Of-Speech(POS) tagging is the essen- tial basis of Natural language process- ing(NLP). In this paper, we present an al- gorithm that combines a variety of context features, e.g. the POS tags of the words next to the word a that needs to be tagged and the context lexical information of a by Canoni- cal Belief Network to together determine the POS tag of a. Experiments on a Chinese cor- pus are conducted to compare our algorithm with the standard HMM-based POS tagging and the POS tagging software ICTCLAS3.0. The experimental results show that our algo- rithm is more effective.