Answer Exaction Of Question Answering Based On Web

Along with the high speed development of internet, the information on the web is more and more..It is a hot research that how to find out the needed information quickly and exactly from very large amount of information source in general domain. At present, the internet not only has the abundant information source, but also the information is opened. How to retrieve the needed information the most of quickly, the most of intelligence and the most of exactly based on the each user have very large amount of information source is the new generation of search engine to develop. Thereby, Question Answering is produced.Question Answering depends on applying information retrieval, information extraction, and natural language processing (NLP) to answer for given domain independent questions written in natural language exactly and simply.Question answering (QA) systems typically consist roughly of question analysis, document/passage retrieval, and answer selection. There are three core questions of question answering: how to comprehend enough at the stage of question analysis, how to find out the related documents from module of information retrieves, and how to exact the answers from the related documents in the module of answer exaction.We are research on the answer exaction in this paper. We utilize formal concept analysis to research the following two parts: to exact answers from the Frequently Asked Questions and to exact answers from the web. We propose a new approach of conceptual clustering the user queries logs with formal concept analysis (FCA). Due to the log data change daily, our methods could be received better performance by use our clustering method. We use our cluster algorithm that is based on the DSBCAN to cluster the user logs firstly. Then these clusters are established the formal context. We attempt to deal with similar questions/queries according to their contents as well as the document click information (cross-references) in the formal context. We mainly use them to build the better concept lattices. In addition, we proposed that navigation can be used to extract answers from the FAQs.We use Multi-strategy based on the web and corpus. Due to the complexity of the structure in the Question Answering, we proposed that use FCA to exact answers. For the different questions, we use the different strategies to exact. We achieved the answer exaction based on the concept matching. Specially, for the definition questions, we utilize the Collaborative Recommenders to select answers.This paper introduced a new personal meta-search engine MySearch, which is based on formal concept analysis. It extracts user's information implicitly and provides real-time response by re-ranking the results. Re-ranking is done by using concept lattice that is built by user's usage logs and the results of source engine. Lastly, the improve re-rank is returned by the MySearch. Experimental results show our method have a significant improvement on satisfaction degree between the search result and user's requirement.Finally, we introduce the experiment results and the evaluation.