Abstract:

A novel concept of quantitative sentiment was proposed, which means the sentiment score of the review. The sentiment score is  derived from the natural language text of the review. To calculate the quantitative sentiment, the dependency relationships between sentiment words are  discussed. A quantitative sentiment algorithm of the sentiment sentence is  presented. Furthermore, the relationship between quantitative sentiment and spam reviews is  discussed. A series of prediction rules are  established through intuitive observation. In the end, the store reviews are  analyzed by establishing a time series with quantitative sentiment as indicator. The spam reviews are  detected efficiently. Experimental results show that the proposed method  has good detection result and outperform existing methods in detecting sore review spam.

Key words: quantitative sentiment; dependency relationships; store review spam