Source code and Data Repository for FETILDA: Fin-tuned Embeddings of Texts In Long Document Analysis
The Python scripts used for running the experiments in the FETILDA paper are located in two main folders, "FIN10K-experiments-source-code" and "US-bank-experiments-source-code".
In order to replicate our experiments, run the Python scripts with the following command line arguments, like so:
Python filename.py max-length section target-value total-#-of-epochs starting-epoch ending-epoch learning-rate
The saved weights of the best models for FETILDA on the FIN10K dataset are available here: .
The saved weights of the best frozen model for FETILDA on the US bank dataset are available here: .
The csv files containing the FIN10K dataset to be used directly by the Python scripts are available here: .
The csv files containing the US bank dataset to be used directly by the Python scripts are available here: