Weird inflects but OK: Making sense of morphological generation errors

Kyle Gorman; Arya D. McCarthy; Ryan Cotterell; Ekaterina Vylomova; M. Silfverberg; Magdalena Markowska

Weird inflects but OK: Making sense of morphological generation errors

Kyle Gorman
, Arya D. McCarthy
, Ryan Cotterell
, Ekaterina Vylomova
, M. Silfverberg
, Magdalena Markowska

Language Technology

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

Abstract

We conduct a manual error analysis of the CoNLL-SIGMORPHON 2017 Shared Task on Morphological Reinflection. In this task, systems are given a word in citation form (e.g., hug) and asked to produce the corresponding inflected form (e.g., the simple past hugged). This design lets us analyze errors much like we might analyze children's production errors. We propose an error taxonomy and use it to annotate errors made by the top two systems across twelve languages. Many of the observed errors are related to inflectional patterns sensitive to inherent linguistic properties such as animacy or affect; many others are failures to predict truly unpredictable inflectional behaviors. We also find nearly one quarter of the residual "errors" reflect errors in the gold data. © 2019 Association for Computational Linguistics.

Original language	English
Title of host publication	CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference
Editors	Mohit Bansal, Aline Villavicencio
Number of pages	12
Place of Publication	Stroudsburg, PA
Publisher	The Association for Computational Linguistics
Publication date	2019
Pages	140-151
ISBN (Electronic)	978-1-950737-72-7
Publication status	Published - 2019
MoE publication type	A4 Article in conference proceedings
Event	Conference on Computational Natural Language Learning - Hong Kong, Hong Kong Duration: 3 Nov 2019 → 4 Nov 2019 Conference number: 23 https://www.conll.org/2019

Fields of Science

Animacy
Error taxonomy
Linguistic properties
Errors
113 Computer and information sciences
6121 Languages

Access to Document

weird inflectsFinal published version, 441 KBLicence: CC BY

https://www.aclweb.org/anthology/K19-1.pdfLicence: CC BY

Cite this

Gorman, K., McCarthy, A. D., Cotterell, R., Vylomova, E., Silfverberg, M., & Markowska, M. (2019). Weird inflects but OK: Making sense of morphological generation errors. In M. Bansal, & A. Villavicencio (Eds.), CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference (pp. 140-151). The Association for Computational Linguistics. https://www.aclweb.org/anthology/K19-1.pdf

@inproceedings{6f336244680441dc922be1e149dedd02,

title = "Weird inflects but OK: Making sense of morphological generation errors",

abstract = "We conduct a manual error analysis of the CoNLL-SIGMORPHON 2017 Shared Task on Morphological Reinflection. In this task, systems are given a word in citation form (e.g., hug) and asked to produce the corresponding inflected form (e.g., the simple past hugged). This design lets us analyze errors much like we might analyze children's production errors. We propose an error taxonomy and use it to annotate errors made by the top two systems across twelve languages. Many of the observed errors are related to inflectional patterns sensitive to inherent linguistic properties such as animacy or affect; many others are failures to predict truly unpredictable inflectional behaviors. We also find nearly one quarter of the residual {"}errors{"} reflect errors in the gold data. {\textcopyright} 2019 Association for Computational Linguistics.",

keywords = "Animacy, Error taxonomy, Linguistic properties, Errors, 113 Computer and information sciences, 6121 Languages",

author = "Kyle Gorman and McCarthy, \{Arya D.\} and Ryan Cotterell and Ekaterina Vylomova and M. Silfverberg and Magdalena Markowska",

year = "2019",

language = "English",

pages = "140--151",

editor = "Mohit Bansal and \{Villavicencio \}, Aline",

booktitle = "CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference",

publisher = "The Association for Computational Linguistics",

address = "United States",

note = "Conference on Computational Natural Language Learning, CoNLL 2019 ; Conference date: 03-11-2019 Through 04-11-2019",

url = "https://www.conll.org/2019",

}

Gorman, K, McCarthy, AD, Cotterell, R, Vylomova, E, Silfverberg, M & Markowska, M 2019, Weird inflects but OK: Making sense of morphological generation errors. in M Bansal & A Villavicencio (eds), CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference. The Association for Computational Linguistics, Stroudsburg, PA, pp. 140-151, Conference on Computational Natural Language Learning, Hong Kong, Hong Kong, 03/11/2019. <https://www.aclweb.org/anthology/K19-1.pdf>

Weird inflects but OK: Making sense of morphological generation errors. / Gorman, Kyle; McCarthy, Arya D.; Cotterell, Ryan et al.
CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference. ed. / Mohit Bansal; Aline Villavicencio . Stroudsburg, PA: The Association for Computational Linguistics, 2019. p. 140-151.

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

TY - GEN

T1 - Weird inflects but OK

T2 - Conference on Computational Natural Language Learning

AU - Gorman, Kyle

AU - McCarthy, Arya D.

AU - Cotterell, Ryan

AU - Vylomova, Ekaterina

AU - Silfverberg, M.

AU - Markowska, Magdalena

N1 - Conference code: 23

PY - 2019

Y1 - 2019

N2 - We conduct a manual error analysis of the CoNLL-SIGMORPHON 2017 Shared Task on Morphological Reinflection. In this task, systems are given a word in citation form (e.g., hug) and asked to produce the corresponding inflected form (e.g., the simple past hugged). This design lets us analyze errors much like we might analyze children's production errors. We propose an error taxonomy and use it to annotate errors made by the top two systems across twelve languages. Many of the observed errors are related to inflectional patterns sensitive to inherent linguistic properties such as animacy or affect; many others are failures to predict truly unpredictable inflectional behaviors. We also find nearly one quarter of the residual "errors" reflect errors in the gold data. © 2019 Association for Computational Linguistics.

AB - We conduct a manual error analysis of the CoNLL-SIGMORPHON 2017 Shared Task on Morphological Reinflection. In this task, systems are given a word in citation form (e.g., hug) and asked to produce the corresponding inflected form (e.g., the simple past hugged). This design lets us analyze errors much like we might analyze children's production errors. We propose an error taxonomy and use it to annotate errors made by the top two systems across twelve languages. Many of the observed errors are related to inflectional patterns sensitive to inherent linguistic properties such as animacy or affect; many others are failures to predict truly unpredictable inflectional behaviors. We also find nearly one quarter of the residual "errors" reflect errors in the gold data. © 2019 Association for Computational Linguistics.

KW - Animacy

KW - Error taxonomy

KW - Linguistic properties

KW - Errors

KW - 113 Computer and information sciences

KW - 6121 Languages

M3 - Conference contribution

SP - 140

EP - 151

BT - CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference

A2 - Bansal, Mohit

A2 - Villavicencio , Aline

PB - The Association for Computational Linguistics

CY - Stroudsburg, PA

Y2 - 3 November 2019 through 4 November 2019

ER -

Gorman K, McCarthy AD, Cotterell R, Vylomova E, Silfverberg M, Markowska M. Weird inflects but OK: Making sense of morphological generation errors. In Bansal M, Villavicencio A, editors, CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference. Stroudsburg, PA: The Association for Computational Linguistics. 2019. p. 140-151