Extracting product features from online Chinese reviews

Because of the uneven quality of online reviews, the accuracy of product feature extraction from Chinese reviews is not satisfied. For this reason, we propose a method based on the traditional FP-Growth algorithm and Word2Vec model to extract product features from online Chinese reviews in the clothing field. This paper has two contributions. One is to add semantic similarity calculation to avoid low-frequency feature words being deleted in the first step of FP-Growth algorithm. The other is to construct semantic rules to extract latent product features, which makes up the deficiency of the traditional association rule algorithm. An experiment is run for the data set of Chinese reviews on clothing products, which shows that the proposed method can improve the accuracy rate without affecting the recall rate.