FinCausal 2023 (FinCausal-23)
Inter-sentence (whole document/section) No Polarity Signal No Strength Signal
FinCausal 2023 annotates cause/effect spans in excerpts from financial disclosures and news, continuing FinCausal 2020's shared task with an expanded, re-annotated dataset.
Unlike FinCausal 2020, this release's public data is entirely pre-filtered causal segments — there are no non-causal examples at all, so causality-detection is not listed as a supported task above: a detection table built from this dataset alone would be single-class (every row Causal) and meaningless to train or evaluate on in isolation. The conversion script still writes causality-detection/{train,test}.parquet anyway (derived from the identification table via causalatee.data.utils.identification_batch_to_detection) so it's there for anyone who wants to pool it with another dataset's negatives for a combined detection table — it's just not advertised as a first-class supported task here, and this project's own evaluation sweep excludes it. The official host is a CodaLab competition gated behind shared-task registration; this project instead converts from a participant's public mirror of the labeled training data, since the organizers never released the equivalent of FinCausal 2020's plain download — see the conversion script for the exact provenance and the train/test split this project had to invent (no split exists in the source beyond an unlabeled blind test).
Overview
| Domain | Financial news |
| Year | 2023 |
| Sentences | 2,630 |
| Granularity | Inter-sentence (whole document/section) |
Tasks & Splits
| Task | Train | Test |
|---|---|---|
| Causal Event Candidate Extraction | — | — |
| Causality Identification | — | — |
| Causality Extraction | — | — |
Citation
@inproceedings{moreno-sandoval:2023,
title = {The {{Financial Document Causality Detection Shared Task}} ({{FinCausal}} 2023)},
booktitle = {2023 {{IEEE International Conference}} on {{Big Data}} ({{BigData}})},
author = {{Moreno-Sandoval}, Antonio and {Porta-Zamorano}, Jordi and {Carbajo-Coronado}, Blanca and Samy, Doaa and Mariko, Dominique and {El-Haj}, Mahmoud},
year = 2023,
month = dec,
pages = {2855--2860},
doi = {10.1109/BigData59044.2023.10386745},
url = {https://ieeexplore.ieee.org/document/10386745/},
urldate = {2025-04-14},
keywords = {Big Data,causality detection,Conferences,financial documents,NLP,Object recognition,Task analysis}
}