EventStoryLine (ESL)

Intra-sentence Polarity Documented, Not Converted No Strength Signal

EventStoryLine annotates causal and temporal relations between events in 58 topics of ECB+ news clusters, targeting cross-document event coreference and storyline extraction.

The underlying ECB+ event markables distinguish negated event mentions (e.g. NEG_ACTION_OCCURRENCE) from their affirmative counterparts, but this converter currently ingests both undifferentiated, so that polarity signal is documented here without yet being surfaced in the converted relations.

causal-candidate-extraction is derived from the identification table above (causalatee.data.utils.identification_batch_to_extraction), restricted to the entities backing an actual causal relation — ESL2HF's direct ECB+ XML parser only implements detection and identification.

Overview

Domain	News
Year	2017
Sentences	2,247
Granularity	Intra-sentence

Tasks & Splits

Task	Train	Test
Causality Detection	2,014	233
Causal Event Candidate Extraction	780	84
Causality Identification	2,014	233
Causality Extraction	—	—

Data Explorer

Causality Detection

Causal Event Candidate Extraction

Causality Identification

Citation

@inproceedings{caselli:2017,
  title      = {The {{Event StoryLine Corpus}}: {{A New Benchmark}} for {{Causal}} and {{Temporal Relation Extraction}}},
  shorttitle = {The {{Event StoryLine Corpus}}},
  booktitle  = {Proceedings of the {{Events}} and {{Stories}} in the {{News Workshop}}@{{ACL}} 2017, {{Vancouver}}, {{Canada}}, {{August}} 4, 2017},
  author     = {Caselli, Tommaso and Vossen, Piek},
  editor     = {Caselli, Tommaso and Miller, Ben and van Erp, Marieke and Vossen, Piek and Palmer, Martha and Hovy, Eduard H. and Mitamura, Teruko and Caswell, David},
  year       = 2017,
  pages      = {77--86},
  publisher  = {Association for Computational Linguistics},
  doi        = {10.18653/V1/W17-2711},
  urldate    = {2024-07-22}
}