CASEWorkshopTimeLineExtraction

Repository containing the code and data for the 'Timeline Extraction from Decision Letters Using ChatGPT' to appear in the 2024 CASE Workshop.

Directory structure

Below is a short description of the structure of the repository.

• InterAnnotatorAgreement/: Contains the annotations and notebook for the calculation of the IAA scores.
  • InterAnnotatorAgreement.ipynb: Notebook containing the IAA calculation for the dates, events and relations.
• data/: Contains the dataset and precalculated dataframes with the spacy part of the algoritm already run.
  • txt_files Original text extracted from PDF documents, with all sentences present
  • ground_truth_data: dataset with all triples, with the dates manually corrected and converted to ISO format.
  • annotated_triples : dataset with the triples and the output of the date correction algorithm.
  • uncorrected_spacy_output: The output of the spacy algorithm for date extraction pre-run on the train and test sets.
• scripts/
• TimeLineExperiments.ipynb/: Notebook with the code integrated to run the complete pipeline and to evaluate the various components.

Running the experiments

• To run the experiments you can run the 'TimeLineExperiments.ipynb' notebook, which will guide you through the various parts of the algorithm.