Research on the attribute classification of sentiment target based on the stratified sampling
Hanxiao Shi, Xiaojun Li, Hong Liu, Liu-qing Zhu · 2016
Fine-grained sentiment analysis is a hot research topic in the research area of natural language processing and information extraction, including the reorganization of sentiment elements, the determination of sentiment polarity and so on. In order to further improve the performance of the fine-grained sentiment analysis, this paper mainly studies the classification method on attributes of the target. We propose a semi-supervised learning method into the research of attribute classification to reduce the dependence of tagged corpus so as to overcome the difficulties on annotation work on fine-grained sentiment tagged corpus. First, this paper studies the stratified sampling model and bootstrapping method. Second, this paper designs the Seed selection algorithm based on entropy computation of the attribute classification probability. At last, this paper studies the problem of expansion of train corpus comparing the random method with stratified method. The result verifies the bootstrapping based on the stratified sampling model has better performance.