DocumentCode :
2231197
Title :
A new polarity clustering algorithm based on semantic criterion function for text of the Chinese commentary
Author :
Xu, Bin ; Zhang, Yufeng
Author_Institution :
Res. Center of Inf. Resources, Wuhan Univ., Wuhan, China
Volume :
4
fYear :
2010
fDate :
20-22 Aug. 2010
Abstract :
The mining methods for comment text polarity are usually used to adopted supervised learning algorithms, but supervised learning algorithms require significant manual labor marked the training set, and its text set in dealing with will be also faced with dimension disaster, sparse vector, high spatial and temporal complexity, low recall and precision rates that cannot be used for a flood of text polarity classification task. In response to these circumstances, this article will introduce a new polarity clustering algorithm for text of the Chinese commentary, constructed specifically for the Chinese comment on the polarity of the text polarities dictionary meaning of words, a criterion function based on semantic means clustering K-means algorithm. The study is the use of semantic clustering method based on Chinese texts deal with a subjective exploration. The methodologies of experiment, statistics, and analysis are used to do this research. The results of experiment showed that average recall rate of 81.22%, average accuracy rate of 67.76%, indicating that the algorithm is feasible and effective.
Keywords :
data mining; learning (artificial intelligence); pattern clustering; statistics; text analysis; Chinese commentary; dimension disaster; polarity clustering algorithm; semantic criterion function; semantic means clustering K-means algorithm; statistics; supervised learning algorithms; temporal complexity; text polarity classification task; Semantics; algorithm; criterion function; polar Semantic Dictionary; text clustering;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Advanced Computer Theory and Engineering (ICACTE), 2010 3rd International Conference on
Conference_Location :
Chengdu
ISSN :
2154-7491
Print_ISBN :
978-1-4244-6539-2
Type :
conf
DOI :
10.1109/ICACTE.2010.5579668
Filename :
5579668
Link To Document :
بازگشت