Research On The Recognition Of Chinese Comparative Sentences For Product Reviews

As a kind of expression with a strong emotional tendency,comparative sentences are often used to compare two or more things to show people’s preference for different things.Therefore,using natural language processing technology to identify comparative sentences from product reviews on the Internet can effectively obtain customer feedback on products.In order to accurately identify the comparative sentence,it is necessary to perform a linguistic analysis of the comparative sentence first,using words as the basic unit to redefine the comparative sentence into five parts: comparative subject,comparative attribute,comparative characteristic word,comparative object and comparative result word.Simple sentence structure composed.Secondly,according to the difference of the comparative characteristic words,the comparative sentences are divided into two categories: explicit comparative sentences and implicit comparative sentences.On this basis,we summarize and establish a comparative sentence recognition mechanism with words and parts of speech as the rule elements.Finally,based on the relevant characteristics of natural language processing technology,a production rule recognition method based on multi-feature fusion can be proposed.This method first uses the established domain product name list and comparative vocabulary list to perform word segmentation and part-of-speech tagging on the corpus,and then conduct experiments through the production rule recognition system.The experimental data results show that the method has a better recognition effect,and the recognition effect of different rules is also different.The rule with the best recognition effect also means that the comparative sentence represented by the rule is used most frequently.