Recognition of Comparative Sentences from Online Reviews Based on Multi-feature Item Combinations

At present, comparative sentences in online reviews are a common and convincing expression. In the autonomous recognition of Chinese comparative sentences, the selection of feature items plays a important role. The previous research mainly adopt the pattern recognition methods. This paper focuses on the recognition of comparative sentences for multi-feature item combinations in online reviews and use the text classification algorithm in machine learning to achieve. First, analyze the influence of the number of different feature items in comparative sentence recognition about the classification performance, and select the number of feature items with the highest mean of classification accuracy, make a combination of different feature items. Then use the document frequency method to reduce the dimension of feature items and select the Boolean weights to construct feature vector. Finally, using SVM classifier to discern comparative sentences. Based on the online reviews of mobile phone, This paper studies the recognition of comparative sentences for thirty feature items.