Research On Personalized Information Retrieval Modelbased On Neural Network

In the field of information retrieval, it's difficult to learn user's interest by user behavior. But neural network model can combine user behavior features at the hidden layer by activation function and form the more abstract representation at the top layer, then learn user's interest. On the basis of the existing artificial neural network model, this paper focused on the study of neural network model in the field of personalized information retrieval, specifically the use of neural networks' relevant characteristics, automatically learning user interest and establishing a link between the search terms and the user's interest. The main work can be summarized as follows:Firstly, for the problem that the user's behavior is often ignored, this paper presents a personalized information retrieval model based on neural networks. Through the comparison of multi group experiments, it is proved that the neural network model can automatically learn the user's interests and demonstrate the influence of the number of hidden layer and the number of hidden layer nodes on the personalized information retrieval results..Secondly, the neural network models can learn user's interest feature automatically by user behavior, but it ignores the important role of connectedness between words and documents for information retrieval results. Aiming at the problem, this paper presents a personalized information retrieval method based on model fusion. Considering the two aspects of user's interest and the relevance between search term and documents, this method integrates the BM25 model and neural network model and meets the actual needs of personalized information retrieval. Experimental results show that the model based on the fusion of the proposed method is superior to other similar personalized retrieval algorithms.Finally, this paper implements a system of the personalized information retrieval based on the neural network model. The system shows the results achieved by our proposed method in the personalized information retrieval domain by concrete examples.