Research Article
Sentiment Analysis of Chinese Words Using Word Embedding and Sentiment Morpheme Matching
@INPROCEEDINGS{10.1007/978-3-030-00916-8_1, author={Jianwei Niu and Mingsheng Sun and Shasha Mo}, title={Sentiment Analysis of Chinese Words Using Word Embedding and Sentiment Morpheme Matching}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 13th International Conference, CollaborateCom 2017, Edinburgh, UK, December 11--13, 2017, Proceedings}, proceedings_a={COLLABORATECOM}, year={2018}, month={10}, keywords={Sentiment polarity analysis Word formation rule Sentiment morpheme matching Word embedding Synonym expanding}, doi={10.1007/978-3-030-00916-8_1} }- Jianwei Niu
Mingsheng Sun
Shasha Mo
Year: 2018
Sentiment Analysis of Chinese Words Using Word Embedding and Sentiment Morpheme Matching
COLLABORATECOM
Springer
DOI: 10.1007/978-3-030-00916-8_1
Abstract
Sentiment analysis has become significantly important with the increasing demand of Natural Language Processing (NLP). A novel Chinese Sentiment Words Polarity (CSWP) analyzing method, which is based on sentiment morpheme matching method and word embedding method, is proposed in this paper. In the CSWP, the sentiment morpheme matching method is creatively combined with existing word embedding method, it not only successfully retained the advantages of flexibility and timeliness of the unsupervised methods, but also improved the performance of the original word embedding method. Firstly, the CSWP uses word embedding method to calculate the polarity score for candidate sentiment words, then the sentiment morpheme matching method is applied to make further analysis for the polarity of words. Finally, to deal with the low recognition ratio in the sentiment morpheme matching method, a synonym expanding step is added into the morpheme matching method, which can significantly improve the recognition ratio of the sentiment morpheme matching method. The performance of CSWP is evaluated through extensive experiments on 20000 users’ comments. Experimental results show that the proposed CSWP method has achieved a desirable performance when compared with other two baseline methods.
