论文总字数：18870字

摘 要

随着网络的发展,人们的生活方式也在不断变化,各个方面也都在陆续的创新和进步。持续的改变使得我们不得不在一些技术上逐步跟进,才能更好地贴近的这个遍地网络化的时代。网络不仅是实现人们互相友好交流，表达情感,分享生活,轻松交友、发表看法等得最佳平台,也是现在人们获取信息的重要来源之一。网络中充满着大量不同用户发布的不同文本信息，或是对各个城市各种美食的点评，或是对新闻报道社会事件的个人看法,或是对所遇事情各种情感的充分表达,或是对世间万物各处风景的赞美,等等等等。充分利用这些数据，能为各个应用领域提供大量强而有力的帮助。但是，再这些数据每天以爆炸式的是速度疯狂增长的同时，也带来了巨大的问题,依旧采用传统的人工方式进行处理。陈旧的处理方法的效率无法再满足数据的飞速增长带来的问题,而且往往时间一长,人工的处理结果就显得不那么准确。因此，情感分析技术营运而生，它可以完全替代传统的人工方式,精确而批量的处理数据,工作效率不受时间的长短而限制,可以自动分析文本中的态度信息、情感信息,做出分类判断,从而替代人工方式,大量解放劳动力。

通过查阅大量的文献和数据,本文对目前文本情感分析领域的研究现状进行了详细调查和统计分析，并对其中存在的明显问题和不足进行了研究,提出了一些仅代表自己的个人看法。其主要研究内容如下：

1. 提出了一种基于词向量的判断方法，从数据库寻找与该词语的近义词,进行分类判断,在一定程度上解决了不完全和不正确的问题。
2. 提出了一种结合语义规则的取向方法。讨论情感的转移是否会影响剧情感极性,以及含有多种不同态度句子的分析是否也遵循这个规则,检验此种方法的适用程度,提高分析准确性。
3. 用基于递归神经网络的方法对文本句子和文本文件进行情感多分类分析，从而判断出文本的情感倾向。

关键词：情感倾向性分析； 词向量； 句法分析； 语义规则； 递归神经网络

Sentiment analysis of short text based on deep learning

Abstract

With the rapid development of Internet technology, the network is not only an important source for people to obtain information, but also gradually becomes the best platform for people to express their personal views and share their own experiences. "Reading network" has gradually changed into "interactive network". The Internet is full of massive text information released by users themselves. These texts are either the users' independent evaluation of a product or service, or the public's personal views on a news event or national policy. If we can make full use of these data, we can provide help for government public opinion monitoring, enterprise market analysis and other application fields. However, these data grow exponentially every day. It takes a lot of manpower and time to mine and analyze them manually. How can we get the information we need from the massive network resources by effective means? Text sentiment analysis technology came into being.

Text sentiment analysis is one of the hot topics in the field of computer research. It aims to automatically analyze the attitude information in the text through computer technology, and make a classification judgment of the emotional tendency of the text. This paper investigates and analyzes the research status of Chinese text sentiment analysis, aiming at the existing problems and deficiencies in this research field, the main research contents and innovations include:

(1) Aiming at the difficulty of sentiment analysis caused by the popularity and use of Internet words, this paper puts forward a method to judge the sentiment tendency of words based on word vector. With the help of the training of language model, the word vector is used to represent the words, and the similarity between words is mined from the corpus itself. According to the similarity of the word vector, the emotional tendency of unknown words is judged, which effectively solves the problems of imperfection and difficult expansion of the traditional emotional dictionary.

(2) combined with the word vector based sentiment analysis method in the first part, a Chinese sentence sentiment orientation analysis method is proposed, which combines syntax tree and semantic rules. On the one hand, the syntactic information of the sentence is fully considered, and the analysis of the sentence is realized on the syntactic tree structure. On the other hand, aiming at the emotional polarity transfer phenomenon of the sentence, combined with the negative words, degree words, turning words and other characteristics that affect the emotional polarity of the sentence, the corresponding semantic rules are defined, which effectively improves the accuracy of the sentence emotional tendency analysis.

(3) According to the characteristics of text sentence emotion diversification and text involving a wide range of fields, this paper uses the method based on recurrent neural network to analyze text sentence and text file emotion multi classification, so as to judge the emotional tendency of the text.

Keywords: Emotional orientation analysis; Word vector; Syntactic analysis; Semantic rules; Recurrent neural network

目录

摘要 I

Abstract II

第一章 绪论 1

1.1. 论文工作背景和意义 1

1.2. 相关研究现状 1

1.3. 主要工作内容 4

1.4. 论文组织结构 4

第二章 相关的基本工具和技术基础 5

2.1. 基本工具 5

2.1.1. Pycharm开发工具介绍 5

2.1.2. Webstorm介绍 5

2.1.3. Python语言介绍 5

2.1.4. SnowNLP类库介绍 5

2.2. 基本技术 5

2.2.1. FastApi、vue以及开发过程使用到的配置文档介绍 5

2.2.2. Uvicorn介绍 6

2.2.3. Vue框架的优势介绍 6

2.3. 本章小结 6

第三章 系统的需求分析及可行性研究 8

3.1. 系统功能需求分析 8

3.1.1. 系统数据流程分析 8

3.1.2. 文本情感分析模块 8

3.1.3. 文件上传模块 9

3.1.4. 文件文件情感词汇提取模块 9

3.1.5. 文件文件情感分析模块 9

3.2. 系统非功能需求分析 9

3.2.1. 性能需求 9

3.2.2. 环境需求 10

3.3. 本章小结 10

第四章 系统设计 11

4.1. 系统总体设计 11

4.1.1. 系统结构设计 11

4.1.2. 系统功能模块结构设计 11

4.1.3. 系统总体工作流程设计 12

4.2. 系统详细设计 12

4.3. 本章小结 13

第五章 系统具体实现 14

5.1. 系统实现工具与环境 14

5.2. 核心代码分析 14

5.3. 系统主要运行界面 17

5.3.1. 文本情感分析模块 17

剩余内容已隐藏，请支付后下载全文，论文总字数：18870字