AMUSE – financial summarization


Train new model:

python –task train –start 0 –count
–modelpath data/models/new_model.h5 –train data/train –gold data/gold

data/train = dir where the text files are data/gold = dir where the gold summaries are

Trains new AMUSE prediction model for given files and stores it in an .h5 file

Generate summaries with existing model:

python –task generate-summaries –start 0 –count
–modelpath data/models/new_model.h5 –test data/test/ –summarydir data/summaries

Also stored:

a model trained on 3000 files named

If you use this code, please cite:

Litvak M, Vanetik N. Summarization of financial reports with AMUSE. In Proceedings of the 3rd Financial Narrative Processing Workshop 2021 (pp. 31-36).

@inproceedings{litvak2021summarization, title={Summarization of financial reports with AMUSE}, author={Litvak, Marina and Vanetik, Natalia}, booktitle={Proceedings of the 3rd Financial Narrative Processing Workshop}, pages={31–36}, year={2021} }


View Github