摘要
文章基于提取用户评论中用户的观点和评论的极性,利用基于句法分析的模式匹配方法来提取特征词,建立特征词集合。采用了TF-IDF算法来抽取匹配特征词的观点词集合,去除贡献度低的观点词,然后采用简约相似度算法来量化种子词和目标观点词之间的相似度,该相似度能够很好的代表未知目标词在一个句子中的倾向性,从而为量化用户观点极性打下基础。最后以某电商平台上的商品成交记录作为实验测试数据,得到不同商品在不同特征之间的评价差异,从而挖掘出用户的个性化需求。
The paper takes the trading records of water heaters on the business platform as an example,extracts the viewpoints and polarity of reviews based on the users＇ comments,and proposes feature words based on the matching pattern of syntactic analysis so as to establish sets by data collection and pretreatment. The paper also extracts view words sets of matching feature words by TF-IDF algorithm,and get rid of those view words which contributes less. Then the simple similarity algorithm is used to quantify the similarity between seed words and target view words. The similarity can be very good to the tendentiousness of the unknown target words in a sentence,laying a solid foundation for quantifying users＇ views polarity. Finally the evaluation of differences among different features of different water heaters can be seen,and the users＇ individualized requirements can be mined.
出处
《广东石油化工学院学报》
2016年第1期35-39,共5页
Journal of Guangdong University of Petrochemical Technology
基金
广东省高等学校科技创新项目(2013kjcx0132)
国家级大学生创新创业训练计划项目(201411656017)
校级大学生创新创业训练与培育项目(2015DCA004
2015py A002
2015py A041
2015py A042)
大学生拔尖创新人才培养"培英计划"项目(广石化院[2015]21号)
关键词
评论极性
产品评论挖掘
用户观点抽取
Polarity of reviews
Product reviews mining
Users＇ view extraction
