Law Recommendation Based on Self - Attention Mechanism and Feature Fusion

Abstract

Aiming at the problem of insufficient understanding of features and time-consuming calculation of long short-term memory networks in convolutional neural networks when extracting features, and the problem that both of them cannot reflect the importance of each word in the whole when extracting features, a method of law recommendation based on self-attention mechanism and feature fusion is proposed. Firstly, the text is preprocessed and Word2vec is used for word vectorization. Then the BIGRU model is used to extract the context features of the text, and the self-attention mechanism is added to extract the weighted information after the BIGRU features are extracted. CNN model is used to extract local features of text; finally, the characteristics of attention mechanism and CNN are fused to effectively solve the problems existing in a single model. The experimental results of the data set from the Judiciary Artificial Intelligence Challenge of China Law Research Cup show that the proposed model is better than the single model and its improved model.