Micro-blogging Opinion Mining Using Co-Occurrence Chains

detection of micro-blogging in the area of opinion mining. Firstly, we parse messages into message-chains by taking advantages of the explicit reply marks in micro-blogging. Then message-chains are clustered into different features (or facets) by comparing the degree of co-occurrence among them. After that, we perform sentiment analysis using semantic-based SBV polarity algorithm. We also proposed two heuristics according to the specificities of micro-blogging. Experimental evaluations show that heuristic co-occurrence chain based algorithm can extract discriminative and meaningful features and outperform those methods we previously proposed. The main research areas are as follows:1. Large - scale distributed crawler technique for Micro-blogging. We design and achieve large-scale distributed crawler so that it can efficiently and rapidly collect and obtain the corpus of some topic from Micro-blogging.2. Web pages based metadata analysis technique. Using special templates with high performance and easy to expand, we extract metadata form HTMLs and form messages to the formation of message-chains.3. Micro-blogging opinion mining using co-occurrence chains. Combining the technologies of Topic Model in TDT (Topic detection and tracking), messages chains are formed into co-occurrence chains. And we can easily analysis the orientation of the features (or facets) of topics.4. Design and realize the prototype system of Sentiment Analysis for SINA Micro-blogging. Using the provided API, large-scale distributed crawler technologies and the heuristic co-occurrence chains algorithm, topic features or facets can easily be found and then by analyzed for their orientation.