Optimization of Feature-Opinion Pairs in Chinese Customer Reviews

Customer reviews mining can urge manufacturers to improve product quality and guide people a rational consumption. The commonly used mining methods are not satisfactory in precision of the features and opinions extracting. In this paper, we extracted the product features and opinion words in a unified process with semi-supervised learning algorithm, and made an adjustment of the threshold value of confidence to obtain a better mining performance, then adjusted the features sequence with big standard deviation, and maximized the harmonic-mean to raise the precision while ensured the recall. The experiment results show that our techniques are very effective.