Analysis Of Short Text Sentiment Orientation For Evaluation Objects

The Analysis of user's emotional tendency based on Online user reviews Refers to the use of machine learning methods or semantic analysis preprocessing,classification,clustering,feature extraction and sentiment analysis on user comments,such subjective text information,so as to effectively excavate the potential value information in the text,such as the user's position,emotion,or improvement advice on a certain object.Although the study of emotional tendency of the comment text has achieved certain results,most remain in the commendatory and derogatory emotion research of the whole text,lack of emotional orientation research for evaluation objects.But the user needs the fine grained emotion analysis for the evaluation object,so this research has some practical significance.On the basis of previous studies on Natural Language Processing,an emotional tendency analysis model for evaluating objects is proposed.In this paper,we use the depth learning tool to obtain the word vector,and the word vector is incorporated into the graph model,which combined with the Syntactic dependence relationship to extracted emotional objects and emotional features separately,and improved of Ku's method to calculate the emotional value of the evaluation feature,that's improved the accuracy of the evaluation of the characteristics of the emotion.The main work of this paper is divided into the following parts according to the process of reviewing text processing.(1)We constructed a set of emotional analysis models extracted from objects and evaluation words,which made the sentiment analysis of the reviews more fine-grained.(2)In the aspect of evaluation object extraction,we use the similarity degree based on word2 vec to improve the edge weights of the traditional TextRank,and use it to extract the evaluation objects.(3)In the aspect of evaluating feature extraction,we use emotion dictionary.Based on syntactic dependency relations,the rule of extraction of evaluation words is formulated,and use it completed the extraction of evaluation features.(4)With the help of the idea of TF-IDF,the emotional intensity of Ku's emotional words was improved,and the combination of evaluation objects and evaluation features was analyzed.