In the reviews, customers use emotional words to express their own views about the products. Consequently, mining the emotional words automatically become one of the hot spots in reviews mining. For the emotional words abstraction of Chinese product reviews, we approach from the two aspects of part of speech and syntax in this paper. We mine the part of speech templates of emotional words from training set firstly, then extract emotion words set based on the templates, and prune the emotional words set through the dependent relationship and stop words set at last. The final extraction results not only include the emotional words which evaluate the dominant features, but also include the ones about the invisible features and overall evaluations. The method is not limited to certain categories of products, so it has the extensive applicability. The experimental results show the effectiveness of this method.
展开▼