中国电子商务平台产品评论意见挖掘——基于条件随机场模型的实证研究

OPINION MINING OF CHINESE E-COMMERCE PLATFORM PRODUCTS COMMENTS—EMPIRICAL RESEARCH BASED ON CONDITIONAL RANDOM FIELD

电子商务行业已成为国家战略性新兴行业,不仅拉动中国经济增长,更改变了人们的生活方式.对电子商务平台产品评论的意见信息进行统计分析,对于了解消费者对产品的关注点,改善平台购物体验,促使生产商对产品改进升级等具有重要意义.互联网时代,数据类型已从单一的结构化数据扩展到文本、图片等非结构化数据.文本挖掘是对大量非结构化数据处理和分析的过程.意见挖掘在文本挖掘基础上添加了人工智能,可以更有效地分析文本数据中的意见信息.文章以京东商城魅族MX3的用户评论为基础数据,采用意见挖掘中的条件随机场模型,并且在模型中加入了是否评价句特征,提高了条件随机场模型的绩效,通过对比试验验证了特征的有效性,从而对意见信息进行分类和可视化分析.

The e-commercial industry has become a strategic emergent industry of a country. To learn customers＇ opinion about the product, e-commerce websites carry on customer management. The abstraction of opinion information mainly adopts opinion mining technology. And it is developed from data mining, to which artificial intelligence is added on the basis of data mining. The article takes the comments of a product made by users on the e-commerce platform websites as the objects of study, and gets the products＇ characteristic which users care about as well as their comments by adopting users＇ comments on the e-commerce platform and analysis these opinions. At first, we preproeess the corpus and annotate it. Then we adopt conditional random fields model added the characteristic of whether evaluation to identify the opinion information. Finally we add up the opinion information of the corpus and extract users＇ opinion from the test corpus through collecting evaluation objects and sentiment words. We find characteristics of the product in which customers are interested and what need to improve. Later we put forward some policy suggestions according to these conclusions.