为实现网络上非领域相关的中文产品评论信息的搜索及结构化,设计并实现中文产品评论结构化引擎,该引擎通过抓取网络上指定产品的评论,抽取评论中的特征与意见,判断意见的极型,将评论以(特征,意见,极型)的结构化形式输出给用户。其中特征意见抽取解决了产品隐形特征抽取问题。实验结果表明该引擎的特征意见召回率与准确率、意见极型判断准确率等指标可以满足用户需要。
We designed and implemented a structured system of mining Chinese product reviews .According to the user ’ s speci-fied product names , the system could grab reviews about the product on Internet , extract product features and automatically deter-mine the polarity of context-dependent opinions and then return a structured result of the reviews as the triple (feature, opinion, polarity).The experiment proved the effectiveness of the algorithms .And it made an acceptable precision and recall .