Research On Recommendation Algorithm Based On LDA And Its Application In Literature Retrieval

With the development of IT and the Internet, the network information facing humanity is showing explosive growth. So how to block out useless information from a large number of text information and get the target information, has been a hot topic in the study of natural language processing problem. A fundamental problem in the current text processing field presence, is how to quantify the characteristics described in the text. The pros and cons of text feature is largely determined by calculating the similarity of text directly, which is cornerstone of the clustering, the recommendation and content-based approach. So the theme of this paper uses LDA model to quantify the characteristics of descriptive text, combines with the potential inherent in the text itself subject information to improve the accuracy of the text similarity calculation. And in the field of library literature retrieval to carry out the application research based on the LDA model retrieval method.The main work of this paper includes the following two aspects:First, this paper proposes a text similarity calculation method based on the theme LDA model. The method by using the LDA model for modeling the text and using Gibbs sampling of MCMC method, works out the distribution of the text and topic also the distribution of the theme and the key words. Further we calculate the similarity between the text based on the keyword weight distribution of the reference topic model. This calculation method is the foundation of the recommended algorithm as follow.Second, in order to improve the quality of text recommendation we introduce the LDA topic model to handle the problem of the text recommendation. Using the proposed text characteristic description method based on the LDA as similarity calculation of measurements improves the content-based recommendation algorithm. Then it return the highest similarity Top-N recommendation as a result, so as to improve the text of the recommended quality.Third, we design a prototype based on the LDA model retrieval system. This system has the characteristics of high cohesion and low coupling. The system uses an event-driven based on Listener- Runner structure, which makes the retrieval system has the characteristics of asynchronous process.