摘要
指出情感标签由评价对象和情感词组成,包含评论的关键要素,能清楚地表达评价者的观点意见。提出一种针对产品网络评论的情感标签抽取模型,利用依存句法分析设计情感标签抽取算法,通过情感极性计算对抽取出的情感标签进行过滤。通过放宽的抽取规则与情感极性过滤相结合,以提高情感标签的召回率,实现潜在评价对象的抽取。最后用网络抓取的产品评论语料作为测试数据集对模型进行测试,获得较高的抽取准确率和召回率,并对模型中存在的问题进行总结,作为模型改善的指导。
As a collection of evaluation object and sentiment words, sentiment label contains key elements of on user reviews and can effectively reflect their core contents. This paper proposes a model of extracting sentiment label from products reviews in Web. Based on dependency parsing technology, it designs the algorithm of sentiment label extraction, and filters extracted sentiment labels by setting the threshold for emotional polarity. Combined with the relaxed rules of extraction and emotional polarity filter, it gets higher recall and extracts the potential target of reviews. Finally, it captures on- line reviews of product as the test data set to test the model and receive a higher precision and recall. To improve the model, the problems in the model are also summarized.
出处
《图书情报工作》
CSSCI
北大核心
2014年第14期12-20,共9页
Library and Information Service
基金
国家自然科学基金项目“科研团队动态演化规律研究”(项目编号:71273196)
北京市财政项目“大数据环境下情报服务规范化体系建设”(项目编号:PXM2013_178214_000010)
武汉大学自主科研项目(人文社会科学)“网络视角下的应急情报体系建设主题研究”(项目编号:274014,得到“中央高校基本科研业务费专项资金”资助)的研究成果之一
关键词
情感标签
观点挖掘
依存句法分析
产品评论
sentiment label
opinion mining
dependency parsing
product review
