Protest-er: Retraining bert for protest event extraction

Tommaso Caselli; O. Mutlu; Angelo Basile; A. Hürriyetoğlu

doi:10.18653/v1/2021.case-1.4

Protest-er: Retraining bert for protest event extraction

Tommaso Caselli, O. Mutlu, Angelo Basile, A. Hürriyetoğlu

Computational Linguistics (CL)

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

7 Citations (Scopus)

84 Downloads (Pure)

Abstract

We analyze the effect of further retraining BERT with different domain specific data as an unsupervised domain adaptation strategy for event extraction. Portability of event extraction models is particularly challenging, with large performance drops affecting data on the same text genres (eg, news). We present PROTEST-ER, a retrained BERT model for protest event extraction. PROTEST-ER outperforms a corresponding generic BERT on out-of-domain data of 8.1 points. Our best performing models reach 51.91-46.39 F1 across both domains.

Original language	English
Title of host publication	Proceedings of the 4th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2021)
Editors	Ali Hürriyetoğlu
Publisher	Association for Computational Linguistics (ACL)
Pages	12-19
Number of pages	8
DOIs	https://doi.org/10.18653/v1/2021.case-1.4
Publication status	Published - 2021

Access to Document

10.18653/v1/2021.case-1.4Licence: CC BY

Protest-er: Retraining bert for protest event extractionFinal publisher's version, 421 KBLicence: CC BY

Handle.net

Persistent link

Cite this

Caselli, T., Mutlu, O., Basile, A., & Hürriyetoğlu, A. (2021). Protest-er: Retraining bert for protest event extraction. In A. Hürriyetoğlu (Ed.), Proceedings of the 4th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2021) (pp. 12-19). Association for Computational Linguistics (ACL). https://doi.org/10.18653/v1/2021.case-1.4

@inproceedings{9e8e8099c4f14c2d9d860d39ff733f95,

title = "Protest-er: Retraining bert for protest event extraction",

abstract = "We analyze the effect of further retraining BERT with different domain specific data as an unsupervised domain adaptation strategy for event extraction. Portability of event extraction models is particularly challenging, with large performance drops affecting data on the same text genres (eg, news). We present PROTEST-ER, a retrained BERT model for protest event extraction. PROTEST-ER outperforms a corresponding generic BERT on out-of-domain data of 8.1 points. Our best performing models reach 51.91-46.39 F1 across both domains.",

author = "Tommaso Caselli and O. Mutlu and Angelo Basile and A. H{\"u}rriyetoğlu",

year = "2021",

doi = "10.18653/v1/2021.case-1.4",

language = "English",

pages = "12--19",

editor = "Ali H{\"u}rriyetoğlu",

booktitle = "Proceedings of the 4th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2021)",

publisher = "Association for Computational Linguistics (ACL)",

}

Caselli, T, Mutlu, O, Basile, A & Hürriyetoğlu, A 2021, Protest-er: Retraining bert for protest event extraction. in A Hürriyetoğlu (ed.), Proceedings of the 4th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2021). Association for Computational Linguistics (ACL), pp. 12-19. https://doi.org/10.18653/v1/2021.case-1.4

Protest-er: Retraining bert for protest event extraction. / Caselli, Tommaso; Mutlu, O.; Basile, Angelo et al.
Proceedings of the 4th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2021). ed. / Ali Hürriyetoğlu. Association for Computational Linguistics (ACL), 2021. p. 12-19.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Protest-er

T2 - Retraining bert for protest event extraction

AU - Caselli, Tommaso

AU - Mutlu, O.

AU - Basile, Angelo

AU - Hürriyetoğlu, A.

PY - 2021

Y1 - 2021

N2 - We analyze the effect of further retraining BERT with different domain specific data as an unsupervised domain adaptation strategy for event extraction. Portability of event extraction models is particularly challenging, with large performance drops affecting data on the same text genres (eg, news). We present PROTEST-ER, a retrained BERT model for protest event extraction. PROTEST-ER outperforms a corresponding generic BERT on out-of-domain data of 8.1 points. Our best performing models reach 51.91-46.39 F1 across both domains.

AB - We analyze the effect of further retraining BERT with different domain specific data as an unsupervised domain adaptation strategy for event extraction. Portability of event extraction models is particularly challenging, with large performance drops affecting data on the same text genres (eg, news). We present PROTEST-ER, a retrained BERT model for protest event extraction. PROTEST-ER outperforms a corresponding generic BERT on out-of-domain data of 8.1 points. Our best performing models reach 51.91-46.39 F1 across both domains.

U2 - 10.18653/v1/2021.case-1.4

DO - 10.18653/v1/2021.case-1.4

M3 - Conference contribution

SP - 12

EP - 19

BT - Proceedings of the 4th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2021)

A2 - Hürriyetoğlu, Ali

PB - Association for Computational Linguistics (ACL)

ER -

Protest-er: Retraining bert for protest event extraction

Abstract

Access to Document

Handle.net

Fingerprint

Cite this