TUdatalib Upgrade

Am 2. Juni erfolgte ein TUdatalib Upgrade auf eine neue Softwareversion. Dieses Upgrade bringt wichtige Neuerungen mit sich. Eine Übersicht finden Sie in der Dokumentation
On June 2nd, TUdatalib was upgraded to a new software version. This upgrade introduced major changes to the system. Please see our documentation for an overview.

 
Open Access

Football Coreference Corpus

Thumbnail Image

Files

FCC_FCC-T_ECBp_dates_2022-05-25.zip () (910.79 KB)

Date

2020-03-19

Type

Dataset
Text

Authors

Bugert, Michael
Reimers, NilsORCID 0000-0002-8663-1131
Barhom, Shany
Dagan, Ido
Gurevych, Iryna

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Description

This script generates: 1. the original sentence-level Football Coreference Corpus (*FCC*), 2. a version of the sentence-level FCC which was cleaned and updated after manual review, 3. *FCC-T*, the extended version of the Football Coreference Corpus with reannotated token-level spans, 4. and publication date annotations for the *ECB+* corpus. The script downloads the original documents from archive.org's WaybackMachine, cleans and processes them locally on your machine and combines the result with our annotations. See README.md for instructions. Details on the annotations and corpora: * For the original FCC, see Bugert et al. 2020 "Breaking the Subtopic Barrier in Cross-Document Event Coreference Resolution", http://ceur-ws.org/Vol-2593/paper3.pdf * For the token-level reannotation FCC-T, see Bugert et al. 2020 "Cross-Document Event Coreference Resolution Beyond Corpus-Tailored Systems", https://arxiv.org/abs/2011.12249 In case of trouble with this downloader, please get in touch on Github: https://github.com/UKPLab/cdcr-beyond-corpus-tailored/issues Cross-document event coreference resolution (CDCR) is the task of detecting and clustering mentions of events across a set of documents. A major bottleneck in CDCR is a lack of appropriate datasets, which stems from the difficulty of annotating data for this task. We present the first scalable approach for annotating cross-subtopic event coreference links, a highly valuable but rarely occurring type of cross-document link. The annotation of these links requires combing through hundreds of documents - an endeavor for which conventional token-level annotation schemes with trained expert annotators are too expensive. We instead propose crowdsourcing annotation on sentence level to achieve scalability. We apply our approach to create the Football Coreference Corpus (FCC), a corpus of 451 sports news reports, while reaching high agreement between NLP experts and crowd annotators in the process.

Keywords

Citation

Identifier

https://tudatalib.ulb.tu-darmstadt.de/handle/tudatalib/2305

Endorsement

DFG Classification

4.43-04 Künstliche Intelligenz und Maschinelle Lernverfahren
4.43-05 Bild- und Sprachverarbeitung, Computergraphik und Visualisierung, Human Computer Interaction, Ubiquitous und Wearable Computing

Project(s)

Faculty

Collections

Coreference Resolution

License

Except where otherwise noted, this license is described as Creative Commons Attribution Share-Alike 4.0