基于深度神经网络的微博短文本情感分析研究

发布时间：2018-08-24 16:50

【摘要】：近年来,随着社交网络的逐渐成熟和移动终端技术的迅猛发展,微博作为一种网络传播的主要媒体形式,越来越受到人们的青睐。用户通过在微博上表达观点传播思想,抒发个人情感的同时,也产生了大量带有个人主观情感特征的信息,这些信息中包含着不同趋向的情感特征,进而对网络舆情的传播能产生巨大的影响。本文使用深度学习的方法,对互联网上微博短文本的情感分析问题进行了相关研究。具体研究内容如下:(1)为了更好的判定微博短文本的情感极性,提出一种基于深度卷积神经网络模型的情感分类方法。该方法首先将训练的词向量作为原始特征向量,然后把特征向量送入卷积神经网络(CNNs,Convolutional Neural Networks)模型进一步提取特征,训练出基于该网络的情感分类模型,再使用该分类器对互联网短文本进行情感分类。实验比较了基于传统机器学习的SVM算法与深度学习的随机生成向量的CNNs模型法和本文提出的方法,最终通过实验结果证明了采用本文方法可以有效的进行情感分类。(2)针对微博短文本中评价对象抽取的问题,提出了一种基于双向长短时记忆循环神经网络(Bidirectional Long Short-Term Memory,BLSTM)模型的情感要素抽取方法。通过实验对比传统机器学习模型与循环神经网络(RNN,Recurrent Neural Networks)、长短时循环神经记忆网络(Long Short-Term Memory,LSTM)和双向长短时记忆循环神经网络这三种深度学习模型发现,采用基于深度学习的双向长短时记忆循环神经网络模型处理评价对象抽取任务可以获得最佳效果。
[Abstract]:In recent years, with the gradual maturity of social network and the rapid development of mobile terminal technology, Weibo, as the main media form of network communication, is more and more popular. By expressing their views on Weibo, users spread their ideas, expressing their personal feelings, and at the same time, they also produced a large amount of information with the characteristics of personal subjective emotions, which contain different trends of emotional characteristics. In turn, the spread of network public opinion can have a huge impact. This paper studies the affective analysis of Weibo's short text on the Internet by using the method of deep learning. The main contents are as follows: (1) in order to better judge the emotional polarity of Weibo's short text, a method of emotion classification based on deep convolution neural network model is proposed. Firstly, the trained word vector is taken as the original feature vector, and then the feature vector is sent into convolution neural network (CNNs,Convolutional Neural Networks) model) to extract features, and then the emotion classification model based on this network is trained. Then we use the classifier to classify the Internet text. The experiment compares the CNNs model method based on the traditional machine learning algorithm with the CNNs model method of the random generating vector of the depth learning and the method proposed in this paper. Finally, the experimental results show that this method can be used to effectively classify emotion. (2) aiming at the problem of evaluation object extraction in Weibo's short text, An emotional element extraction method based on bidirectional long and short term memory loop neural network (Bidirectional Long Short-Term Memory,BLSTM) model is proposed. Compared with traditional machine learning model, cyclic neural network (Long Short-Term Memory,LSTM) and bidirectional long and short term memory circulatory neural network (Long Short-Term Memory,LSTM), we find that, Bidirectional long and short time memory loop neural network model based on deep learning can be used to deal with the evaluation object extraction task and the best result can be obtained.
【学位授予单位】：中原工学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1

【参考文献】