# Extracting Fine-Grained Economic Events from Business News Source code and experiment replication data for the proceedings paper "Jacobs G. and Hoste V. 2020. Extracting Fine-Grained Economic Events from Business News. FNP-FNS @ COLING28. Forked from https://github.com/dwadden/dygiepp @4f77e4d5703facbc5edf4535d065899975e84e90 Read original documentation at that commit for detailed setup and instructions. ## Replication data: - For SENTiVENT economic event data: Preprocessed files in `data/sentivent/`. - For ACE05_Event follow instructions on how to obtain in `DYGIEPP_README.md` (requires LDC licensing). - Best trained Pytorch weights model for SENTiVENT in `models/sentivent-event-nonerforargs/`. - Spreadsheet overviews of all hyperparameter runs in `predictions/`. ## Running the event pipeline for experiments - To build image: `docker build --tag dygiepp:dev .` - To run image with all gpus: `docker run --gpus all -it --ipc=host --name dygiepp dygiepp:dev` - To run with specific device: `docker run --gpus "device=2" -it --ipc=host --name dygiepp dygiepp:dev` - To run for dev mount volume: `docker run --gpus all -it --ipc=host -v /home/gilles/repos/dygiepp/:/home/gilles/repos/dygiepp/ --name dygiepptrainsentivent dygiepp` - `cd /dygiepp && conda init bash` - `exec bash` - `conda activate dygiepp` Start training: `rm -rf ./models/ace05-event; bash ./scripts/train/train_ace05_event.sh 0` - To predict a trained model: `allennlp predict models/<MODEL_DIR>>/model.tar.gz data/sentivent/ner_with_subtype_args/test.jsonl --predictor dygie --include-package dygie --use-dataset-reader --output-file ./predictions/sentivent-<SETTINGSNAME>-test.jsonl --cuda-device 0` When in container: - To evaluate a trained model: `allennlp evaluate models/<MODEL_DIR>>/model.tar.gz data/sentivent/ner_with_subtype_args/test.jsonl --include-package dygie --output-file predictions/sentivent-<SETTINGSNAME>-metrics_test.jsonl --cuda-device 0` ## Contact - Gilles Jacobs: gilles@jacobsgill.es, gilles.jacobs@ugent.be - Veronique Hoste: veronique.hoste@ugent.be ## Mirrors - https://osf.io/j63h9/ - https://github.com/GillesJ/sentivent-coling-fnp-fns