Automatically predicting the polarity of Chinese adjectives: Not, a bit and a search engine

Ge Xu, Chu-ren Huang, Houfeng Wang

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

1 Citation (Scopus)


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.
Original languageEnglish
Title of host publicationChinese Lexical Semantics - 14th Workshop, CLSW 2013, Revised Selected Papers
Number of pages13
Publication statusPublished - 1 Dec 2013
Event14th Workshop on Chinese Lexical Semantics, CLSW 2013 - Zhengzhou, China
Duration: 10 May 201312 May 2013

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8229 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference14th Workshop on Chinese Lexical Semantics, CLSW 2013


  • Chinese adjective
  • polarity
  • sentiment analysis

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Automatically predicting the polarity of Chinese adjectives: Not, a bit and a search engine'. Together they form a unique fingerprint.

Cite this