Entity

Time filter

Source Type


Yang Y.,Shandong Normal University | Yang Y.,Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology Shandong Province | Liu P.,Shandong Normal University | Liu P.,Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology Shandong Province | And 4 more authors.
ICIC Express Letters | Year: 2013

The sentiment analysis of phrases occupies important positions among the multiple application domains of sentiment analysis. The previous work of the sentiment analysis of phrases is still dependent on sentiment words. However, context-dependent problem of sentiment words, which refers that the same sentiment word indicates different sentiment orientations with respect to different contexts, is a challenging task on sentiment analysis. In order to resolve this problem, this paper proposes a method for sentiment analysis based on potential sentiment expectation of aspects. In this method, we define sentiment expectation of aspects, and construct the relationship among aspects, sentiment words and phrases. On the basis of this relationship, we construct a model to analyze the orientation of phrases. Experiment results show that our method can effectively address the dependence of sentiment words on context, and enhance the degree of precision for sentiment analysis of evaluation phrases. © 2013 ICIC International.


Fei S.,Shandong Normal University | Fei S.,Shandong University of Finance and Economics | Liu P.,Shandong Normal University | Liu P.,Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology Shandong Province | And 4 more authors.
ICIC Express Letters | Year: 2013

Aiming at the problem of topic correlated detection caused by the fact that one topic contains many subtopics, this paper proposes a subtopic partitioning method based on the improved ant colony clustering algorithm. First of all, this method establishes a loss function which is used to evaluate the loss degree of clustering and overcome the problem of semantic irrelevancy caused by the cosine similarity of ant colony clustering algorithm. Then this method carries out semantic clustering to the report by using the improved algorithm for the purpose of subtopic partitioning. At last, this method realizes the correlated detection of the topic by using the relative entropy. The experimental result indicates that the method described in this paper is an effective subtopic partitioning method. © 2013 ICIC International.

Discover hidden collaborations