A Web entity activity recognition approach based on k-nearest neighbors classifier

Based on the traditional information extraction, this paper puts forward an approach to recognizing entity activity form Web. Entity activity which describes the behaviour or action information about the entity is valuable to recognize for search engine, business intelligence, data integration and so on. Since Web entity activities, as multi-relationships, are contained in unstructured text in the Web, recognizing entity activities is a difficult problem in current information extraction area. In this paper, we rigorously define the recognition task firstly and propose a novel recognition framework to extract entity activities form Web. Firstly, this paper proposes a k-nearest neighbors classifier based on sentence similarity to discover the sentences which contain web entity activity. Then, according to dependency parsing feature, we use a set of heuristic rules to extract the information from sentences. Experimental results demonstrate the feasibility and effectiveness of our proposed approach and it can adapt to multi-domains.