Open Nav

基于深度学习的文本情感分析研究

以下是资料介绍,如需要完整的请充值下载.
1.无需注册登录,支付后按照提示操作即可获取该资料.
2.资料以网页介绍的为准,下载后不会有水印.仅供学习参考之用.
   帮助中心
资料介绍:

基于深度学习的文本情感分析研究(论文11000字)
摘        要
随着网络的急速发展,大量网络平台涌现。在社交平台上如微博、微信、抖音,人们可以讨论对热点事件的看法。在游戏平台如Steam、Wegame,有大量用户发表对游戏的看法。在电子商务平台如拼多多、苏宁等有大量的客户发表对产品的评价。对于政府部门而言,能够充分了解公众对热点事件的态度和把握社会舆情。对于游戏制作者、电子商务卖家,能够及时充分了解用户对产品的满意程度,有助于提高产品质量。所以,研究文本情感分析的方法具有重要的商业价值和社会意义。
虽然循环神经网络和卷积神经网络在文本分类中有着不错的分类效果,但是梯度消失和梯度爆炸是循环神经网络难以解决的问题,而卷积神经网络也存在缺乏利用上下文信息的缺点,因此本文研究结合双向门控循环神经网络和卷积神经网络,实现对文本在句子级别上的情感分类。结合二者的优势,本文提出了一种新的文本分类模型:BIGRU-CNN文本分类模型,并且通过实验验证了这种模型的有效性。
本文在IMDB、微博评论电影评论的数据集上进行了多组对比实验。实验结果表明:BIGRU-CNN模型在微博评论的数据集上取得了92.49%的F1值,比CNN、CNN-BIGRU、BIGRU、BIGRU-ATT模型取得的F1值分别高出了0.24%、0.26%、0.48%、0.6%;BIGRU-CNN模型在IMDB电影评论的数据集上取得了92.51%的F1值,比CNN、CNN-BIGRU、BIGRU、BIGRU-ATT模型取得F1值分别高出了1.18%、1.45%、0.88%、0.61%,这充分说明了本文模型的的有效性。

[资料来源:https://www.doc163.com]


关键词:情感分析    双向门控循环神经网络卷积神经网络    BIGRU-CNN

Research of Text Sentiment Analysis Based on Deep Learning
Zhou Zuguang
(College of Software Engineering, South China Agricultural University, Guangzhou 510642,China)
Abstract:With the fast development of the network, a lot of network platforms have showed. On social platforms such as Weibo, WeChat, and Tik Tok, people can discuss their views on hot events. On gaming platforms such as Steam and Wegame, a large number of users have expressed their opinions on the game.On e-commerce platforms such as Pinduoduo and Suning, many customer have published comments on the products.For government departments, they can fully understand the public's attitude towards hot events and grasp public opinion. For game producers and e-commerce sellers, being able to fully understand the user's satisfaction with the product in a timely manner helps improve product quality. Therefore,The method of studying text classificationplays an essential role in commercial value and social significance.

[资料来源:www.doc163.com]


Although recurrent neural networks and convolutional neural networks have a good classification effect in text classification, the disappearance of gradients and gradient explosions are difficult problems for recurrent neural networks. Convolutional neural networks also have the disadvantage of lacking the use of contextual information, so this article The research combines BIRNN and CNN to realize the sentiment classification of text at the sentence level.A new text classification model: BIGRU-CNN text classification model, is proposed and verified effectiveness of this model through experiments.
In this paper, multiple sets of comparative experiments were conducted on the datasets of Weibo reviews and IMDB movie reviews. The experimental results find that the BIGRU-CNN model achieved an F1 value of 92.49% on the Weibo review dataset, which was 0.24%,0.26%,0.48%,0.6% higher than the F1 valueobtained by the CNN, CNN-BIGRU, BIGRU, and BIGRU-ATT models, respectively. the BIGRU-CNN model achieved an F1 value of 92.51% on the IMDB movie review dataset, which was 1.18%,1.45%,0.88%,0.61% higher than the F1 values obtained by the CNN, CNN-BIGRU, BIGRU, and BIGRU-ATT models, respectively.This demonstrates the effectiveness of the model in this paper.

[资料来源:http://Doc163.com]


Key words: Sentiment Analysis  Bidirectional gated recurrent neural network  Convolutional Neural Network  BIGRU-CNN
 
目        录
1绪论    1
1.1课题研究的背景意义    1
1.2国内外研究现状    1
1.2.1基于情感字典及语义规则的文本情感分析方法    1
1.2.2基于机器学习的文本情感分析方法    1
1.2.3基于深度学习的文本情感分析方法    2
1.3本文的主要内容    2
1.4本文的组织结构    2
2相关技术与理论研究    4
2.1引言    4
2.2词向量    4
2.2.1One-hot 编码    4
2.2.2Word Embedding    4
2.2.3Word2Vec 模型    4
2.3深度学习网络    5
2.3.1卷积神经网络    5
2.3.2循环神经网络    6
2.4本章小结    9
[资料来源:https://www.doc163.com]

3基于深度学习神经网络的文本情感分析    10
3.1引言    10
3.2TextCNN 模型    10
3.3TextRNN 模型    11
3.3.1TextRNN 模型    11
3.3.2引入注意力机制的TextRNN模型    12
3.4BIGRU-CNN网络的分类模型    14
3.5本章小结    15
4实验设计与结果分析    16
4.1实验设计    16
4.1.1实验环境    17
4.1.2实验数据    17
4.2实验结果与分析    18
4.2.1训练阶段    18
4.2.2测试阶段    21
4.2.3提升阶段    22
4.3本章小结    25
5总结与展望    26
参考文献.......... .................27
致谢    29 [版权所有:http://DOC163.com]

  • 关于资料
    提供的资料属本站所有,真实可靠,确保下载的内容与网页资料介绍一致.
  • 如何下载
    提供下载链接或发送至您的邮箱,资料可重复发送,若未收到请联系客服.
  • 疑难帮助
    下载后提供一定的帮助,收到资料后若有疑难问题,可联系客服提供帮助.
  • 关于服务
    确保下载的资料和介绍一致,如核实与资料介绍不符,可申请售后.
  • 资料仅供参考和学习交流之用,请勿做其他非法用途,转载必究,如有侵犯您的权利或有损您的利益,请联系本站,经查实我们会立即进行修正! 版权所有,严禁转载
    doc163.com Copyright © 2012-2024 苏ICP备2021029856号-4