基于深度学习的社交网络平台细粒度情感分析

发布时间：2018-05-22 11:25

本文选题：情感分析 + 深度学习　；参考：《计算机应用研究》2017年03期

【摘要】：在当下互联网迈入Web 2.0时代,多样的社交网络平台呈现出巨量而丰富的文本情感信息,因此挖掘网络数据文本信息并作情感倾向判断对人机交互与人工智能具有重大的现实意义。传统的解决文本情感分析问题的方法主要是浅层学习算法,利用回归、分类等方案实现特征的提取及分类。以这类方法为起点,探索采用深度学习的方法对网络文本进行细粒度的情感分析,以期达到即时获取依附于网络世界的社会人的情感,甚至是让机器达到对人类情感表达的深度理解。对于深度学习的具体实现,采用的是降噪自编码器来对文本进行无标记特征学习并进行情感分类,利用实验训练获得最佳的参数设置,并通过对实验结果的分析和评估论证深度学习对于情感信息的强大解析能力。
[Abstract]:As the Internet enters the Web 2.0 era, a variety of social networking platforms present a huge amount of text emotional information. Therefore, it is of great practical significance for human-computer interaction and artificial intelligence to mine network data text information and make emotional tendency judgment. The traditional method to solve the problem of text affective analysis is the shallow learning algorithm, which uses regression, classification and other schemes to achieve feature extraction and classification. Using such methods as a starting point to explore the use of in-depth learning methods for fine-grained emotional analysis of online texts, with a view to achieving instant access to the emotions of the social people attached to the online world, Even to give the machine a deep understanding of human emotional expression. For the realization of the deep learning, the noise reduction self-encoder is used to learn the unmarked feature of the text and to classify the emotion, and the best parameter setting is obtained by the experiment training. Through the analysis and evaluation of the experimental results, the powerful ability of deep learning to analyze emotional information is demonstrated.
【作者单位】：南京工业大学电气工程与控制科学学院;
【分类号】：TP391.1

【参考文献】