Data and code for the paper "LEDA: a Large-Organization Email-Based Decision-Dialogue-Act Analysis Dataset" at ACL 2023
Instructions:
- We recommend creating a fresh python (3.8.0) environment to run this code. The exact versions of the libraries used can be found in "requirements.txt".
- The data and python notebook to generate the plots and some of the statistics is in the "Notebook" folder, just open it and execute the chunks in order.
- the Notebook will also generate an input file for the dialog tagger, but you can find the same file already precomputed in the "PredictionModel/data" folder
- The code to run the prediction models is in the "PredictionModels/model" folder. From there, execute "./run.sh" and the dev/test set scores for each epoch should be generated in log.txt.
- we recommend running this on GPUs with at least 16GB of RAM (in case of memory errors, reducing the batch size could help)
In case of issues please contact m.karan@qmul.ac.uk.