PRESIDE-lm NER Model

The PRESIDE-lm NER model and weights for detection and extraction of judicial entities and their honorific titles from the text of US federal district court docket sheets. This model was developed on the docket sheet text from cases filed in the 94 US Federal District courts during 2016. Model performance on validation data is 99.3% and 99.2% F-Score for the detection of honorific titles and judge names, respectively.