基于网络的情感语义词典的自动构建技术研究

Research the key technology about Web extraction of semantic emotional concepts and automated construction of semantic lexicon. Using Internet's huge amounts of information and technology, through some given semantic emotional concept of seeds and built lexicons, based on co-occurrence characteristics of retrieval concepts' hits and co-occurrence network structure characteristics of the seeds and extracted concepts and clustering algorithm, to extract the emotional concept in bootstrapping, and to calculate the feelings weights of emotional concepts and the semantic relations' weights between concepts. Discuss the multidimensional structure of emotional semantic dictionary, including multi-stage classification of emotional concept and the building of semantic relations. Building and using derivation semantic rules of emotional relation, to automatic derivation new emotional semantic relations, and to evaluate the semantic lexicon construction and recondition the algorithm for extracting semantic emotional lexicon based the Web. At the same time, to research construction method of Tibetan semantic emotional lexicon based on the constructed Chinese lexicon and Sino-Tibetan lexicon. Finally, based on constructed semantic emotional lexicon in Sino-Tibetan, to study affective computing and analysis of network public opinion, and verify the e effectiveness of semantic emotional lexicon and its building methods based on Web in further.

主要研究基于网络的情感语义词 汇抽取和语义情感词典自动构建的关键技术。利用互联网络海量信息和技术，通过给定的语义情感概念种子和已有情感词典，基于网络信息中情感概念共起互信息频率特征、概念共现网络结构特征以及聚类的算法抽取情感概念，计算情感概念的情感权值和概念间的语义关系权值。讨论多维情感语义词典的构建结构，包括情感概念的多级分类和多种语义关系构建。构建和利用语义关系的推导规则，自动推导新的情感语义关系，并对构建的情感语义词典进行评价和算法修整。同时基于构建的中文情感语义词典和藏汉词汇词典，研究构建藏文情感语义词典。最后通过基于构建获得的汉藏情感语义词典对网络舆情进行细粒度情感计算和情感倾向分析技术研究，进一步验证情感语义词典构建方法和词典的有效性。

主要研究基于网络的情感语义词汇抽取和语义情感词典自动构建的关键技术，以及基于构建的情感语义词典的文本情感分析。.明确了多维情感语义词典的结构，包括情感概念的多级分类和多种语义关系。情感词典包含情感词汇、情感倾向、情感细分类、同反义和行为语义关系等，以及情感词汇的权值和语义关系权值。.利用互联网络海量信息、情感词典和现代汉语词典等资源，基于概念语义关系特征，通过语义关系抽取技术，自动抽取情感词汇和情感语义关系。利用语义关系的推导规则，递归抽取、计算情感词的权值、情感词汇间语义关系及其权值等。自动归类词汇的情感倾向和细化分类。自动抽取情感语义关系，解决了手工构建、标注情感语义词典的费时费力问题。借助地方民族特色和优势，以中文情感语义词典为基础，构建了藏文情感语义词典。构建的汉藏双语情感语义词典，解决了现有汉语情感语义词典语义关系单一、藏文情感语义词典空白的问题。通过与常用情感词典比较，研究实验结果表明有效地抽取了情感词汇和语义关系，扩展了汉藏双语情感词典的语义关系。.基于构建的汉藏情感语义词典对情感计算研究。比较基于不同情感词典的微博情感倾向计算结果，构建的汉藏双语情感语义词典有效地提高了基于情感词典的文本情感倾向分析准确度。利用深度学习算法对藏文微博进行情感倾向分析，针对藏文缺乏微博语料问题，利用藏文情感语义词典同反义词替换来扩充语料数量，解决了深度学习对训练语料数据量的要求，同时提高了藏文情感计算精度。