Research On The Extraction Of Opinion Targets Based On The Part Of Speech Features And Syntactic Analysis

Emotional information extraction mainly includes three basic problems: the extraction of polarity words,the extraction of opinion targets,as well as the extraction of collocation combined with opinion targets and polarity words.According to the words and phrases of fine-grained level in emotional information extraction,selecting the Chinese online comments as a data source,according to the characteristics of Chinese online reviews,such as published initiative by consumer,write at random,unstructured and not have prior knowledge and diversified,the paper proposes a method of opinion target extraction based on the part of speech features and syntactic analysis.Firstly,the rules are constructed by adjective,adverbs,verbs of part of speech feature to extracted polarity words.Secondly,through the syntactic tree structure of clause sequence extract and filter candidate opinion target.Then,based on the core syntactic paths to screen(opinion target,polarity word)on collocation,in order to reduce the noise of opinion target and polarity word introduced in the process of extracting,thus,extract the real opinion target.Finally,in order to extract the opinion targets with more representative,based on Naive Bayesian classification algorithm to classify the opinion targets.Through the experiment of hotel field corpus,in the aspects of accuracy and F Measure are better than the original algorithm.