Extracting and Clustering the Evaluation Objects of the Chinese Product Recommendation System Based on the Opinion Mining
Download as PDF
DOI: 10.23977/iemb.2019.021
Author(s)
Yingbin Xue, Xiaoye Wang, Yingyuan Xiao, Yukun Li, Wenguang Zheng
Corresponding Author
Yingbin Xue
ABSTRACT
The research of product recommendation system mainly focuses on the user s behavior or the commodities. contents, but rarely focuses on the commodies. reviews. This paper extracts useful information hidden in the commodies. reviews by opinion mining technology. It is more targeted that recommending product to users according to the user's favorite property. The main process of opinion mining is the extraction of topic words and the polarity judgement of polar words. Because the time complexity of the topic extracting algorithm is high, this paper extracts the explicit evaluation object and evaluation words by using the method of matching noun phrase and then setting up a semantic mapping set of evaluation objects and evaluation words to determine the implicit evaluation object. In this paper, k-means and BIRCH are combined to cluster the evaluation objects. K-Means algorithm is used for pre-clustering for the BIRCH algorithm to solve local optimum. And the advantage of BIRCH is it can get the number of clusters by self-learning. And delete the clusters contained few contents to pruning evaluation objects. It can reduce the time complexity and guarantees the clustering effect.
KEYWORDS
Product recommendation, opinion mining, extraction of theme words, clustering