摘要： 产品评论挖掘是从自然语言描述的用户评论中获取信息的过程，产品特征抽取是产品评论挖掘的第1个阶段，产品特征的好坏决定了产品评论挖掘中后续阶段的质量。采用弱监督的学习方法，只需要提供少量的产品特征作为种子，从这些种子出现的语句中抽取文本模式，利用文本模式来发现新的产品特征。实验结果表明，从英文文本中自动抽取产品特征的实验系统，取得了较好的效果。

关键词: BootStrapping算法, 文本模式抽取, 产品评论挖掘

Abstract: The mining of product reviews is the process of extracting information in reviews which is expressed by natural language, the extraction of product feature is the first phrase of the mining of product reviews. The quality of product feature decides the quality of subsequent phrases. This paper adopts weakly supervised methods, which just need a hand of product features as the seeds, using the occurrence sentences of seed to extract text patterns, and using the text patterns to find new product features. Experimental results show that it can extract product feature from English plain text and receive good result.

Key words: BootStrapping algorithm, text pattern extraction, product review mining

中图分类号: 

• TP18