Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Text Processing

Research on Micro-Blog Text Sentiment Classification Method and Application

Author KangHao
Tutor WangZuo
School National University of Defense Science and Technology
Course Control Science and Engineering
Keywords Micro-Blog Sentiment Classification SVM Semantic Rules Forecast
CLC TP391.1
Type Master's thesis
Year 2012
Downloads 117
Quotes 0
Download Dissertation

As a very popular online social media, micro-blog is surprisingly booming inrecent years. On micro-blog people can publish a blog or give replies to others’ blogs.Through it people can express opinions, chat with each other, and estimate thepolarity to persons or events. Quickly distinguishing the positive or negative textsfrom all these blogs or replies can be used to reflect the views or position of the publicto the real public figures or social events, while it is also the research point of thispaper. Text sentiment classification is established on accomplishing the classifying ofsentiment texts by using computer instead of the manpower, and is a research hotspotin recent years. The purpose above can be reached by this solution too.We implement micro-blog sentiment classification using two methods, and applyit to a specific instance to reflect the popularity of public figures. Based on the bigdataset collected by our group before, we make the following specific contributions:To solve the problem of the classifying of mico-blog texts, we study thetechniques of micro-blog text sentiment classification using SVM (Support VectorMachines), which includes the selection and expansion of sentiment lexicon, featureselection of sentiment text, and sentiment classification with SVM.But the SVM-based method does not take into consideration the emotionaldivergence. In view of this, we further investigate the sentiment classification basedon semantic rules. We collect a set of semantic rules, construct a sentiment model,and develop the corresponding classification algorithm.Since the sentiment embodied in the micro-blog texts can reflect the popularityof public figures, we apply one of our classification method to a specific instance, andstudy the techniques of forecast based on the sentiment classification.The two methods of micro-blog text sentiment classification we focus on bothget good accuracies which are73.83%and77.17%respectively. We apply thesentiment classification method based on semantic rules to the specific instance, andthen to analysis and forecast. The precison is truly good.

Related Dissertations
More Dissertations