Professional Jurisdiction Recognition for Cross-Domain Filing Based on Deep Hybrid Model

This paper proposes a professional jurisdiction recognition algorithm for case materials in cross-domain filing based on a deep hybrid model. Through the parallel combination of CNN and RNN, the spatial and sequence features of text data can be captured without interfering with each other. In addition, we use the tensor outer product to construct them into a high-order data block with rich information and stronger representation capabilities. Extensive experiments are conducted on a new data set with labeled examples consisting of 2068 case materials from three professional courts and one ordinary courts, and the results demonstrate that the proposed model is effective in professional jurisdiction recognition for cross-domain filing.