Feature selection for Chinese online reviews sentiment classification

Author

Xian Chen ; Jing Ma ; Yueming Lu

Author_Institution

Key Lab. of Trustworthy Distrib. Comput. & Service (BUPT), BUPT, Beijing, China

fYear

2013

fDate

26-28 Oct. 2013

Firstpage

79

Lastpage

82

Abstract

Considering that traditional feature selection methods (DF, MI and IG) usually lost useful information, we propose the Feature Selection for Chinese Online Reviews Sentiment Classification (FSCSC), FSCSC takes empirical analysis into account and focus on how to effectively select different types of features based on statistical approaches to improve sentiment classification performance. FSCSC was tested on a Chinese online reviews corpus with a size of 4000 documents. The experiment indicates that FSCSC can improve the classification effectiveness.

Keywords

behavioural sciences computing; document handling; feature selection; pattern classification; statistical analysis; Chinese online reviews corpus; Chinese online reviews sentiment classification; FSCSC; classification effectiveness; documents; empirical analysis; sentiment classification performance; statistical approaches; traditional feature selection methods; Accuracy; Learning systems; Niobium; Power capacitors; Sentiment analysis; Support vector machines; Text categorization; empirical analysis; feature selection; sentiment classification;

fLanguage

English

Publisher

ieee

Conference_Titel

Computational Problem-solving (ICCP), 2013 International Conference on

Conference_Location

Jiuzhai

Type

conf

DOI

10.1109/ICCPS.2013.6893490

Filename

6893490