An end-to-end joint model for evidence information extraction from court record document

•We explore a new task of information extraction, which aims to extract evidence information from count record documents in the legal domain.•We formalize evidence information extraction task as an integration of paragraph classification and sequence labeling problem, and propose an end-to-end joint model for the task.•Our model achieves the current best performances, outperforming previous methods and neural baseline systems by a large margin.•The proposed model can be applied for better analyzing and understanding legal texts, avoiding a lot of manual work for experts and professionals in the legal domain.