Sentiment Analysis Method Study In The Chinese Microblogging Product Reviews

Customers' consumption satisfaction can be available through questionnaire and telephone follow-up,however,these kinds of sampling survey cannot represent all customers' consumption satisfaction because the sampling data is not typical and representative.Social platform users are quite universal and users' experience information can be stored in large quantity on these platforms.Meanwhile,the micro blog content can easily break the limitations of time,space,ages and occupation,which reflect users' interests,hobbies and emotional thinking in a better way.Therefore,to identify users' interest by acquiring their emotion status and preferences from micro blog comments would provide evidence for merchants to explore potential clients group.This would avail marketing and promotion activities to achieve targeted responses.Micro blog text is quite different from traditional text due to its short,refine and oral-style words.In addition,there are no strict grammar rules in micro blog text;however,this may cause ill-formed sentences.Considering the peculiarity of micro blog text,it shall improve traditional text analysis methods and study suitable emotional analysis methods for Chinese micro blog products and its comments thus to elevate identification accuracy.The thesis' s tasks mainly cover emotional tendency classification,sorting algorithms,retrieval of emotion information.The main study work include the following aspects:To build emotional analysis model of micro blog products comments based on syntactic dependency and textual classification and to compare performance and properties of random forest and SVM support vector machine classification algorithm.To select training sample with high-credibility and micro blog emoji to implement labeling for training sample.To propose emotional analysis algorithm combining syntactic dependency and textual classification.In particular,to have random forest algorithm used in textual classification and conduct experimental analysis including preprocessing,extract of comment words & comment targets and distinguishing emotional tendency thus to analyze emotions in comments for micro blog products.Emotional analysis system for comments on micro blog product is designed.In addition,it expounds the design,research and development process of the system,including design and implementation of data-collection subsystem,data-processing subsystem and data-presenting subsystem.In the end,it tested the system which includes testing environment construction and testing result comparison analysis.Eventually,it identifies the effective emotional analysis model for comments on micro blog products.The emotional analysis algorithm for micro blog in the thesis is built based on Chinese morphology and syntactic analysis.The final emotional analysis system was demonstrated with 88% accuracy of distinguishing subjective and objective sentences and 71% accuracy in distinguishing emotional tendency.