Research And Implementation Of Personalized Agriculture Search Engine Based On The User Interest Model

Currently, agricultural vertical search engines are convenient for users to obtain agricultural information from massive and complex network resources, however the monotonous search results can not satisfy the growing personalized demand of users. They don’t consider the user’s interest, and don’t satisfy the user’s demand for personalized search results. To solve the problems above, this thesis studied the user interest model and personalized sorting algorithms, and realized the personalized agricultural vertical search engine which provides users with personalized information search services. This thesis mainly completed the following studies and work:(1) Study on the construction and update of user interest model. The users’ information is collected from the server logs implicitly, and the interest words are extracted from the user’s historical query records. After combining the weight of interest word in the collection of browse content calculated by the TF-IDF(Term Frequency-Inverse Document Frequency) algorithm with the weight of interest word in the collection of interest words, and this thesis proposes the calculation formula of the interestingness. The user interest model is described by vector space model, and interest attenuation factor is used in the update of the user interest model.(2) Study on personalized sorting algorithm. After analyzing the Lucene sorting algorithm and the ordinary personalized sorting algorithm, this thesis optimizes the common personalized sorting algorithm. Firstly, the optimized personalized sorting algorithm of this thesis uses the constructed user interest model to extend the user’s query before search, secondly, using the cosine formula to calculate the similarity between the user interest model and the search results, so as to simplify the process of similarity calculation. Combining with the Lucene score formula, the comprehensive evaluation formula is proposed. Experimental results show that the personalized sorting algorithm in this thesis is better than the precision of ordinary personalized sorting algorithm.(3) Construction of personalized agricultural search engine. Construction and update method of user interest model, and personalized sorting algorithm ware realized in the agricultural vertical search engine based on Nutch. By using object-oriented programming method, a personalized agricultural search system is constructed. System tests show that it can provide users with personalized search service.