基于词典与机器学习的中文微博情感分析
毕业论文关键词：微博；情感分析；分类；情感倾向加权计算
Chinese Weibo Sentiment Analysis Based on Dictionary and Machine Learning
Abstract：With the rapid development of Internet today, Weibo has become a typical emerging network application. Because of its large number of users, how to utilize the huge data resource of Weibo quickly and effectively becomes a research focus in a variety of fields. This article conducted research on how to obtain users’ emotional orientation from their text on Weibo. Then this article gave a detailed description about the present development of the research on text sentiment analysis and the latest research achievement of Weibo sentiment analysis both at home and abroad, and explained the relative technology and fundamental theory of sentiment analysis of Weibo text in Chinese. After the comparison on the performance of sentiment analysis with different machine learning algorithms and feature combinations, this article designed an emotional orientation weighted algorithm based on emotional dictionary, in order to realize the display system of Weibo sentiment analysis. Finally, the article made further discussions and gave prospects for future development.
Key words: weibo;sentiment analysis;classification; emotional orientation weighted algorithm
目 录
摘要 １
关键词 １
ABSTRACT. １
KEY WORDS １
1.绪论 １
1.1 问题的提出 １
1.2 国内外研究状况 ２
1.3 研究的目的和内容 ２
2 情感分析相关理论与技术 ４
2.1 微博的简介 ４
2.1.1 微博文本的简介 ４
2.2 相关任务的定义 ４
2.2.1 微博主客观的定义 ４
2.2.2 微博极性判别的定义 ５
2.3 情感词典的构建 ５
2.3.1 情感极性词典 ６
2.3.2 副词词典 ６
2.3.3 否定词典 ６
2.3.4 特殊词典 ７
2.4 微博文本预处理 ７
2.4.1 中文分词 ７
2.4.2 去停用词 ８
2.5 特征选择 ８
2.5.1 文档频率（DF） ８
2.5.2 互信息（MI） ８
2.5.3 信息增益（IG） ９
2.5.4 词频（WF） ９
2.5.5 卡方统计量（CHI） ９
2.6 本章小结 ９
3 基于机器学习的微博主客观分类与极性判别 ９
3.1 分类器的基本原理 ９
3.1.1 朴素贝叶斯分类器 １０
3.1.2 支持向量机分类器 １０
3.1.3 逻辑斯蒂回归分类器 １１
3.2 分类器选取 １２
3.2.1 五文非文本特征下的实验 １３
3.2.2 八文非文本特征下的实验 １４