Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/12434
Title: Automatically predicting the polarity of Chinese adjectives: Not, a bit and a search engine
Authors: Xu, G
Huang, C 
Wang, H
Keywords: Chinese adjective
polarity
sentiment analysis
Issue Date: 2013
Publisher: Springer
Source: Lecture notes in computer science (including subseries Lecture notes in artificial intelligence and lecture notes in bioinformatics), 2013, v. 8229 LNAI, p. 453-465 How to cite?
Journal: Lecture notes in computer science (including subseries Lecture notes in artificial intelligence and lecture notes in bioinformatics) 
Abstract: The SO-PMI-IR method proposed by [1] is a simple and effective method for predicting the polarity of words, but it suffers from three limitations: 1) polar paradigm words are selected by intuition; 2) few search engines nowadays officially support the NEAR operator; 3) the NEAR operator considers the co-occurrence within 10 words, which incurs some noises. In this paper, for predicting the polarity of Chinese adjectives automatically, we follow the framework of the SO-PMI-IR method in [1]. However, by using only two polarity indicators, [bu](not) and [youdian](a bit), we overcome all the limitations listed above. To evaluate our method, a test set is constructed from two Chinese human-annotated polarity lexicons. We compare our method with Turney's in details and test our method on different settings. For Chinese adjectives, the performance of our method is satisfying. Furthermore, we perform noise analysis, and the relationship between the magnitude of SO-PMI-IR and accuracy is also analyzed. The results show that our method is more reliable than Turney's method in predicting the polarity of Chinese adjectives.
Description: 14th Workshop on Chinese Lexical Semantics, CLSW 2013, Zhengzhou, 10-12 May 2013
URI: http://hdl.handle.net/10397/12434
ISBN: 9783642451843
ISSN: 0302-9743
EISSN: 1611-3349
DOI: 10.1007/978-3-642-45185-0_48
Appears in Collections:Conference Paper

Access
View full-text via PolyU eLinks SFX Query
Show full item record

Page view(s)

40
Last Week
0
Last month
Checked on Sep 25, 2017

Google ScholarTM

Check

Altmetric



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.