基于条件随机场挖掘产品评论的产品特征词和情感词

Mining product features and opinions from product reviews using conditional random field

为了提高产品评论中产品特征词和情感词的识别率,采用条件随机场(CRF)模型将产品特征词和情感词提取任务转换成词序列标记问题,研究基于条件随机场模型不同特征在抽取任务中的效果,提出了语义角色这一特征并结合了词、词的原形、词性、句子结构信息等特征作为条件随机场模型特征。基于CRF模型使用不同特征进行实验,实验结果验证了语义角色在提取产品特征和情感词应用中起到很好的指示作用。与使用支持向量机模型进行提取任务的实验进行对比,实验结果证明了方法的有效性。

In order to improve the recognition rate of feature words and sentiment words in product reviews, Conditional Random Field（ CRF） model was used to convert the extraction of feature words and sentiment words into word sequence labeling problem, and the effect of different features based on CRF model in the extraction task was studied. The feature of a semantic role was proposed, which combines the features of vocabulary, the prototype, part of speech, sentence structure and so on, for CRF model. Based on the CRF model, the experimental results show that the semantic role if good at extracting product features and emotional words. And compared with the Support Vector Machine（ SVM） model, the experimental results verify the effectiveness of the proposed method.