情感分类研究进展
发布时间:2018-08-30 19:47
【摘要】:文本情感分析是多媒体智能理解的重要问题之一.情感分类是情感分析领域的核心问题,旨在解决评论情感极性的自动判断问题.由于互联网评论数据规模与日俱增,传统基于词典的方法和基于机器学习的方法已经不能很好地处理海量评论的情感分类问题.随着近年来深度学习技术的快速发展,其在大规模文本数据的智能理解上表现出了独特的优势,越来越多的研究人员青睐于使用深度学习技术来解决文本分类问题.主要分为2个部分:1)归纳总结传统情感分类技术,包括基于字典的方法、基于机器学习的方法、两者混合方法、基于弱标注信息的方法以及基于深度学习的方法;2)针对前人情感分类方法的不足,详细介绍所提出的面向情感分类问题的弱监督深度学习框架.此外,还介绍了评论主题提取相关的经典工作.最后,总结了情感分类问题的难点和挑战,并对未来的研究工作进行了展望.
[Abstract]:Text emotion analysis is one of the most important problems in multimedia intelligent understanding. Emotion classification is a core problem in the field of emotion analysis, which aims to solve the problem of automatic judgment of emotional polarity. Due to the increasing scale of Internet comment data, the traditional dictionary-based approach and the machine-learning approach can not deal with the emotional classification of mass reviews. With the rapid development of deep learning technology in recent years, it has shown a unique advantage in the intelligent understanding of large-scale text data. More and more researchers prefer to use depth learning technology to solve text classification problems. It is divided into two parts: 1) generalize and summarize the traditional emotion classification technology, including dictionary-based method, machine-learning based method, and hybrid method. Aiming at the shortcomings of previous affective classification methods, a weak supervised depth learning framework for affective classification problems is introduced in detail. In addition, the classic work related to topic extraction of comments is also introduced. Finally, the difficulties and challenges of affective classification are summarized, and the future research work is prospected.
【作者单位】: 西北大学信息科学与技术学院;
【基金】:国家自然科学基金优秀青年科学基金项目(61522206)~~
【分类号】:TP181;TP391.1
,
本文编号:2214104
[Abstract]:Text emotion analysis is one of the most important problems in multimedia intelligent understanding. Emotion classification is a core problem in the field of emotion analysis, which aims to solve the problem of automatic judgment of emotional polarity. Due to the increasing scale of Internet comment data, the traditional dictionary-based approach and the machine-learning approach can not deal with the emotional classification of mass reviews. With the rapid development of deep learning technology in recent years, it has shown a unique advantage in the intelligent understanding of large-scale text data. More and more researchers prefer to use depth learning technology to solve text classification problems. It is divided into two parts: 1) generalize and summarize the traditional emotion classification technology, including dictionary-based method, machine-learning based method, and hybrid method. Aiming at the shortcomings of previous affective classification methods, a weak supervised depth learning framework for affective classification problems is introduced in detail. In addition, the classic work related to topic extraction of comments is also introduced. Finally, the difficulties and challenges of affective classification are summarized, and the future research work is prospected.
【作者单位】: 西北大学信息科学与技术学院;
【基金】:国家自然科学基金优秀青年科学基金项目(61522206)~~
【分类号】:TP181;TP391.1
,
本文编号:2214104
本文链接:https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2214104.html