AMUSE - financial summarization

Unzip data.zip

Train new model:

python FinAnalyze.py --task train --start 0 --count <how many files,-1 for all> --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 FinAnalyze.py --task generate-summaries --start 0 --count <how many files,-1 for all> --modelpath data/models/new_model.h5 --test data/test/ --summarydir data/summaries

Also stored:

a model trained on 3000 files named model.training.muse.3000.all.h5

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} }