Abstract - Based LDA Retrieval Model

With the development of information technology, social progress, the Internet has gradually become an important part of people’s daily lives. However, the rapid expansion of information resources, causing serious problems of information overload. So how to improve the efficiency of search technology has become a hot issue in the field of information retrieval researchers solved. Traditional search technology can not meet the needs of users of information retrieval. This is the emergence of new search technology provides a strong demand base. Especially in recent years, information retrieval technology innovation, absorbing, constantly improve the effect of a large number of information retrieval methods. By analyzing the former in the field of information retrieval research achievements, additional information on the actual situation of the existence of the text documents from the viewpoint of improvement LDA topic model, the removable Digest method to extract information and introduced into the LDA model proposed Summary-LDA document model, then the model and the traditional probabilistic query retrieval model is proposed by combining the LDA retrieval model based on abstracts. To resolve efficiency of techniques in the current retrieval.The main contents are as follows:First, based on improved LDA model. In theme-based model to study the LDA topic model and its related technologies. For the LDA model does not take full advantage of the presence of a large number of documents such as the author, the issue of additional information highly concentrated full-text abstracts and other content information. This paper presents a digest of information introducing Summary-LDA model, the build process document, the first of extracted digest information modeling, and then use that information to digest document model is smoothed to calculate the probability of a document the most relevant topics, more effective theme extraction capacity.Secondly, LDA retrieval model based on abstract construct. Experimental results show that the presence of traditional query probabilistic retrieval model was unable to dig out text latent semantic information resulting in important missing information and other issues, the paper-based retrieval model based currently retrieval exist by Summary-LDA model superior theme ability to extract the extracted document potential semantic information relating to the integration of the information document relating to traditional query probabilistic retrieval model, abstract model constructed LDA retrieval based on the retrieval process in order to achieve the realization of semantic document information relating to representation, to improve precision retrieval model.Finally, in order to verify the effect of LDA to retrieve abstracts based retrieval model, and the probability of each query retrieval model, clustering and retrieval model based Retrieval Model LDA-based aspect to be compared by calculating precision to retrieve the evaluation results. Experimental results show that the precision LDA-based retrieval model abstracts significantly higher than other models to be improved.