Topology-aware Multi-task Learning Framework for Civil Case Judgment Prediction

Yuquan Le, Sheng Xiao, Zheng Xiao, Kenli Li

Abstract: Highlights•TML adopts PTM to capture semantic interactions between various legal elements.•Assigning special tokens to elements makes TML aware of case structure information.•TML utilizes a multi-task learning paradigm to address multiple-cause situations.•Three parameter-free retrievers are devised to establish inter-task connections.