情感词发现与极性权重自动计算算法研究

发布时间：2018-03-11 13:12

本文选题：情感词　切入点：情感权重　出处：《中文信息学报》2017年03期 　论文类型：期刊论文

【摘要】：随着互联网电子商务和各种社交网络应用的快速发展,产生了大量的用户评价信息。为满足快速整理这些评价信息的需求,情感倾向性分析应运而生。情感词典是各类情感倾向性识别算法的基础,收集一部全面且权重合理的情感词典,往往可以简单快速而有效地解决情感分析问题。但情感词典规模有限,而网络上新的情感词层出不穷,语言使用不规范,人工整理耗时耗力。已有的情感词收集方法较复杂,且领域性强,收集的情感词可扩展性差。本文提出一种自动挖掘潜在情感词并计算其极性权重的算法,该算法与应用领域无关,具有良好的扩展性。该方法利用共现特性,基于朴素贝叶斯公式能检测出未知的情感词,并根据其情感权重值的大小判断其情感极性,可有效地扩展情感词典,将已有的情感词典进一步量化。在理论研究的基础上,本文分别针对京东、豆瓣及大众点评网三组评论语料做了实验,其结果的准确率都基本在90%以上,验证了该方法的有效性和实用性,为情感倾向性分析提供了知识库基础。
[Abstract]:With the rapid development of Internet e-commerce and various social network applications, a large amount of user evaluation information has been generated. Emotion dictionaries are the basis of all kinds of affective orientation recognition algorithms. It is often possible to solve affective analysis problems simply, quickly and effectively. However, the scale of affective dictionary is limited, and new affective words appear in endlessly on the Internet, language usage is not standardized, manual sorting is time-consuming and time-consuming, and the existing methods of collecting affective words are more complicated. This paper presents an algorithm for automatically mining latent affective words and calculating their polar weights, which is independent of the application field and has good expansibility. Based on naive Bayes formula, unknown affective words can be detected, and their affective polarity can be judged according to their emotional weight, which can effectively expand the emotional dictionary and further quantify the existing affective lexicon. In this paper, three groups of comment corpus, JingDong, Douban and Dianping, are tested, and the accuracy of the results is above 90%, which verifies the validity and practicability of this method and provides a knowledge base for emotional orientation analysis.
【作者单位】：北京理工大学计算机学院;北京市海量语言信息处理与云计算应用工程研究中心;公安部第一研究所信安部;工业和信息化部电子科学技术情报研究所;
【基金】：国家重点基础研究发展计划(973计划)(2013CB329601)
【分类号】：TP391.1

【相似文献】