UC-FLAT-based traffic hit-and-run case law document named entity identification method

The invention discloses a traffic hit-and-run case law document named entity identification method based on UC-FLAT. The method comprises the following steps: acquiring annotation data of a plurality of traffic hit-and-run case law documents; extracting legal keywords from the annotated data in an unsupervised manner, and constructing a pre-training dictionary; and a UC-FLAT model is constructed, and entities in the law documents of the traffic hit-and-run case are identified. According to the method, the named entity recognition capability in the aspect of legal instruments is better, proper nouns such as traffic hit-and-run crimes and traffic accidents in the legal instruments can be recognized, and a plurality of entities such as accident types, crime names, primary and secondary responsibilities, sentencing factors, sentencing factors and judgment results can be recognized; the improved unsupervised method is used for constructing the dictionary, word segmentation errors are reduced, better word granulari