论文总字数：38862字

摘 要

随着新时代媒体的快速发展,越来越多的人会在微博、论坛中表明自己的立场、态度和情感，因此网络中曝光的食品安全事件引起更多人的高度关注，用户的评论直接反映着消费者对待食品安全事件的情感倾向。

本文以网络营销理论、电子商务理论、统计学理论等为理论基础，综合运用理论和实证方法，研究食品安全热点事件的情感倾向问题。主要研究内容与结论如下：

首先，分析了食品安全事件的研究背景和意义，对国内外相关文献进行了系统的梳理。

其次，基于食品安全事件的特征，对HowNet词典进行筛选、补充和整合，从基础情感词、程度副词、否定词和停用词四个方面构建了食品安全情感词典。并设计了情感分析的相关指标算法和实验程序。

再次，基于所抓取到的微博用户评论数据，先后对饿了么无证经营黑心作坊、胶水牛排、黄记煌三汁焖锅再曝丑闻这三个近期比较热点的食品安全事件进行了数据挖掘。分别分析了食品安全热点事件的情感倾向和情感倾向的发展趋势，研究结果表明食品安全事件爆发后，人们的负面情感占主导地位且呈下降趋势。进而基于食品安全情感词典，分别计算了三个热点事件的准确率、召回率和F值来对抽取结果进行评价。

最后，本文将三个热点事件的情感特征和情感趋势进行了对比，研究结果表明，舆情的发展主要是由负面情感所决定。食品安全事件的负面评论发展趋势具有一定的共性。即都呈现出迅速下降的趋势，但是也存在一定的不同。同时也将三个事件的准确率、召回率和F值等情感倾向指标进行了对比分析，研究结果表明负面评论信息与词典中的情感词语匹配度高，建立的食品安全情感词典比较有效，能够准确反映食品安全评论的真实情况。

关键词: 食品安全事件 情感词典 情感倾向分析 用户评论

Emotional tendency analysis of food safety hot spot events

ABSTRACT

With the rapid development of new times media, more and more people show their positions, attitudes and emotions in the micro-blog, so food safety incidents in the network exposure is highly concerned by more people, User comments can directly reflect the emotional tendency of consumers towards food safety events.

This paper is based on the theory of network marketing, e-commerce theory, statistical theory , and uses theoretical and empirical methods to study the emotional tendencies of food safety hot events.The main research contents and conclusions are as follows:

Firstly, the research background and significance of food safety events are analyzed, and the relevant literatures at home and abroad are systematically sorted out.

Secondly, based on the characteristics of food safety events, the HowNet thesaurus is screened, supplemented and integrated, and the food safety emotional thesaurus is constructed from four aspects: basic emotion words, degree adverbs, negative words and stop words. And designed the relevant index algorithm and experimental procedure of emotion analysis.

Once again, based on the data of user comments on Weibo, three more recent hot food safety events- the unlicensed black workshops of ‘eleme’, Glue steak, three juice stew pot of Huangjihuang, were data mininged. The trends of emotional tendencies and emotional tendencies of food safety hot events were analyzed respectively. The results showed that people’s negative emotions are dominant after the food safety events and emotional trends are declining. And then based on the food safety emotional thesaurus, respectively, Using the method of calculating the precision, recall rate, and harmonic mean of three hot events separately, then the evaluation results were evaluated.

Finally, this paper compares the emotional characteristics and emotional trends of the three hot events, and the results show that the development of public opinion is mainly determined by negative emotions. The development trend of negative comments on food safety events has some commonalities. The trend shows a rapid decline, but there are some differences. At the same time,this paper makes a comparative analysis of the emotional tendency such as the precision, recall rate, and harmonic mean of the three hot events. The results show that the negative comments are highly matched with the emotional words in the dictionary, and the established emotional dictionary of food safety is effective and can accurate reflect the real situation of food safety reviews.

Keywords: food safety events; emotional dictionary; sentiment analysis; user comment

目 录

摘 要 II

ABSTRACT 1

第一章 绪论 1

1.1 课题来源 1

1.2 研究背景 1

1.2.1 食品安全事件频发 1

1.2.2 微博的快速发展 1

1.3 研究意义 2

1.4 研究内容和研究方法 2

第二章 文献综述 5

2.1 食品安全的相关研究 5

2.2 情感倾向分析的相关研究 6

2.2.1 关于构建情感词典的相关研究 6

2.2.2 情感倾向计算方法的相关研究 7

第三章 情感词典 8

3.1 情感词典简介 8

3.2 食品安全情感词典的构建 9

3.2.1 基本情感词表构建 10

3.2.2 程度词词表构建 12

3.2.3 否定词词表构建 12

3.2.4 停用词词表构建 13

3.3 实验步骤及算法 14

3.3.1实验步骤 14

3.3.2 评价指标规定 15

第四章 情感倾向分析 16

4.1 准备工作 16

4.1.1数据来源 16

4.1.2 数据筛选 16

4.1.3 数据分类 17

4.2 食品安全事件实证结果及分析 23

4.2.1 黑心作坊事件 23

4.2.2“胶水牛排”事件情感倾向分析 25

4.2.3黄记煌三汁焖锅再曝丑闻事件情感倾向分析 28

4.3 本章小结 30

第五章 食品安全热点事件对比分析 32

5.1 三个事件对比分析 32

5.1.1 情感倾向特征对比分析 32

5.1.2 情感趋势对比分析 32

5.2 词典评价对比分析 34

5.3 本章小结 35

第六章 总结与展望 37

6.1 研究结论 37

6.2 研究展望 37

参考文献 39

附录 42

请支付后下载全文，论文总字数：38862字