基于多维扩展特征与深度学习的微博短文本情感分析

发布时间：2018-05-25 05:32

本文选题：社交网络 + 深度信念网络　；参考：《电子与信息学报》2017年09期

【摘要】：该文提出了一种基于深度信念网络(DBN)和多维扩展特征的模型,实现对中文微博短文本的情感分类。为降低传统文本分类方法在处理微博短文时特征稀疏的影响,引入社交关系网络作为扩展特征,依据评论者和博主之间的社交关系,提取相关评论扩展原始微博,将扩展后的多维特征作为深度信念网络的输入。通过叠加多层玻尔兹曼机(RBM)构建DBN模型底层网络结构,多层玻尔兹曼机可以对原始输入抽象并获得数据的深层语义特征。在多个RBM层上叠加一层分类玻尔兹曼机(Class RBM),实现最终情感分类。实验结果表明,通过调整模型参数和网络结构,构建的深度学习模型在情感分类中能够获得比SVM和NB等浅层分类系统更优的结果,另外,实验证明使用扩展多维特征方法可提高短文本情感分类的性能。
[Abstract]:In this paper, a model based on deep belief network (DBN) and multidimensional extended features is proposed to classify Chinese Weibo short texts. In order to reduce the influence of the traditional text classification method on the sparse feature when dealing with Weibo short text, the social relationship network is introduced as the extended feature. According to the social relationship between the reviewer and the blogger, the original Weibo is extracted. The extended multidimensional feature is used as the input of the deep belief network. The bottom structure of the DBN model is constructed by superposing the multi-layer Boltzmann machine. The multi-layer Boltzmann machine can abstract the original input and obtain the deep semantic features of the data. One layer of classification is superimposed on multiple RBM layers to realize the final emotion classification. The experimental results show that, by adjusting the model parameters and network structure, the depth learning model can obtain better results than the shallow classification systems such as SVM and NB. The experimental results show that the extended multi-dimensional feature method can improve the performance of text-based affective classification.
【作者单位】：情感计算与先进智能机器安徽省重点实验室;德岛大学工学部;
【基金】：国家自然科学基金(61432004) 模式识别国家重点实验室开放课题(NLPR)(201407345) 安徽省自然科学基金(1508085 QF119) 中国博士后科学基金(2015M580532)~~
【分类号】：TP391.1;TP393.092

【相似文献】