PMI-IR算法基于无监督学习的情感分析系统

摘要情感观点是人们对事物的一些观点和看法。任何事物都具有两面性，同样情感观点也分正面的和负面的。为了判断对一些事物的情感观点正负性，本文设计了基于无监督方法的情感分析系统，该系统通过互信息算法（PMI-IR）得出情感倾向（SO）得分，然后进行情感分析和文本分类任务。

该系统共分为四个模块，关键短语的提取模块，关键短语搜索模块，互信息算法模块，情感倾向分析模块。关键短语的提取模块是对评论进行词性标注后提取关键词的短语。关键短语搜索模块是对获取的关键词短语进行搜索获取词频。互信息算法模块通过获取的关键短语与基准词共同出现的词频还有基准词单独出现的词频进行的运算。情感倾向分析模块是对所得的SO得分分析判断是否推荐。64603

毕业论文关键词情感分析文本分类观点挖掘 PMI-IR算法

毕业设计说明书（论文）外文摘要

Title A Sentiment Analysis System based on Unsupervised Learning

Abstract Sentiment and the point of view is that people some view of things and ideas. Emotional point of view can be pided into positive and negative. In order to distinguish between good things and bad things, the paper design the Sentiment Analysis System based on Unsupervised learning, which reviews some of the points can be calculated emotional bias score for score for sentiment analysis and text classification.

The system have four modules：The key phrase extraction module, the search of key phrase module, the algorithm of PMI-IR module, and the value of SO analysis modules. The key phrase extraction module is carried out for the review word tagging extracted after two consecutive keyword phrase. The search the key phrase module is for the keyword phrases acquired the word frequency. The algorithm of PMI-IR module through the acquisition of key phrases with the word co-occurrence of word frequency reference and the reference word for word frequency appears alone operation. The value of SO analysis modules obtained for SO values to analysis whether it should recommend.

Keywords Sentiment analysis; Text classification; Opinion mining; The algorithm of the PMI-IR

1 绪论 1

1.1 研究背景 1

1.2 研究现状 2

1.3 本文主要工作 3

2 论文相关工作 4

2.1情感分析任务 4

2.3 本章小结 10

3 系统的设计与实现 10

3.1 系统总体架构设计 11

3.2 关键词短语的提取模块 11

3.2.1 具体任务 11

3.2.2 具体实现 12

3.2.3 功能展示 16

3.3关键短语搜索模块 17

3.3.1 具体任务 17

3.3.2 具体实现 19

3.3.3 功能展示 22

3.4 PMI-IR算法模块 22

3.4.1 具体任务 23

3.4.2 具体实现 23 PMI-IR算法基于无监督学习的情感分析系统:http://www.youerw.com/jisuanji/lunwen_71879.html