| 引用本文: | 谢松县,刘博,王挺.应用语义关系自动构建情感词典.[J].国防科技大学学报,2014,36(3):111-115.[点击复制] |
| XIE Songxian,LIU Bo,WANG Ting.Applying semantic relations to construct sentiment lexicon automaticlly[J].Journal of National University of Defense Technology,2014,36(3):111-115[点击复制] |
|
| |
|
|
| 本文已被：浏览 8546次 下载 7187次 |
| 应用语义关系自动构建情感词典 |
| 谢松县, 刘博, 王挺 |
|
(国防科技大学 计算机学院， 湖南 长沙 410073)
|
| 摘要: |
| 构建英文情感词典研究相对成熟，形成了丰富可靠的词典资源。而针对中文的研究时间短，中文情感分析词典资源较少。借鉴现有可靠的英文词典资源，提出了基于语义关系的情感词典自动构建算法，算法先从HowNet的概念中进行中文义原和词语抽取及语义分析，再利用HowNet概念中DEF中英文属性值，在英文情感词典SentWordNet中进行义原和词语情感值查询，最后根据词语和义原之间的语义关系进行词语的情感值计算。算法直接利用现有的英文情感词典，无须人工标注，生成的情感词典记录了词语的语义关系、情感极性值等多种信息，弥补了现有词典的不足。评测实验结果表明，根据算法实现的情感词典相比其他词典在准确率接近的情况下，召回率和F值最高，取得了较好的评测性能。 |
| 关键词: 情感分析 情感词典 HowNet 语义关系 |
| DOI：10.11887/j.cn.201403020 |
| 投稿日期：2013-10-25 |
| 基金项目:国家自然科学基金资助项目（61170156） |
|
| Applying semantic relations to construct sentiment lexicon automaticlly |
| XIE Songxian, LIU Bo, WANG Ting |
|
(College of Computer, National University of Defense Technology, Changsha 410073, China)
|
| Abstract: |
| Researches on constructing English sentiment lexicon is relatively mature, and there are abundant and reliable lexical resources. Whereas for Chinese studies, the research history is short, and there are only a few Chinese sentiment lexicon resources. With reliable English sentiment lexicon as reference, an automatic constructing approach was proposed, based on semantic relationships. Firstly the Chinese sememe and words were extracted from the defination of concepts in HowNet and the semantic analysis was carried out upon them; secondly the sentimental value of each sememe and word was retrieved from the English sentiment lexicon SentiWordNet according to the DEF attributes of concepts in HowNet, and the final sentimental value of each word was calculated on the semantic relations of the sememe and words. The ready English lexicon was used without manual labeling in the method, and diverse information of words was recorded in the final lexicon, including semantic relations and sentimental values, which remedy the lack of other lexicons. The experimental results show that the resulted sentiment lexicon can achieve better performance in the recall and F value measurements under the condition of approaching other lexicons on the precision measurements. |
| Keywords: sentiment analysis sentiment lexicon HowNet semantic relation |
|
|
