Sentiment analysis aims to predict sentiment tendency automatically. Traditional methods tackling this problem are mostly based on supervised learning, but it is time-consuming and uneasy to extendable. In this paper, we provide a novel method of sentiment analysis based on unsupervised learning together with some language rules. It is no necessary to have a positive sentiment dictionary beforehand as we can build it automatically during processing the comments. By this positive sentiment dictionary, it provides an efficient way to classify the product reviews. The methodology presented is easy to extend due to its un-domain-dependency. As we can see, the experiment result obtained shows its promising application.
展开▼
