Linguistic Enrichment of Historical Dutch using Deep Learning

Silke Creten; Peter Dekker; Vincent Vandeghinste

Linguistic Enrichment of Historical Dutch using Deep Learning

Silke Creten, Peter Dekker, Vincent Vandeghinste

Onderzoeksoutput: Article › peer review

49 Downloads (Pure)

Samenvatting

This article discusses the automatic linguistic enrichment of historical Dutch corpora through the use of part-of-speech tagging and lemmatization. Such a type of enrichment facilitates linguistic research where manual annotation is unfeasible. We built a neural network-based model using the PIE framework and performed an in-depth error analysis, in order to identify the strengths and weaknesses of each approach with respect to labeling historical data. In order to do so, we experimented with two data sets: the Corpus Gysseling (13th century texts) and the Corpus van Reenen/Mulder (14th century texts). We used two different statistical approaches (MBT and HunPos) as baselines for our neural approach. MBT is a memory-based tagger frequently used for modern Dutch, while HunPos is an open source trigram tagger. We present thoroughly analyzed results. In general, the neural model scores better than the two baselines, even with limited training data. Based on the error analysis, we propose several strategies for future research in order to improve the labeling of historical Dutch.

Originele taal-2	English
Pagina's (van-tot)	57-72
Aantal pagina's	16
Tijdschrift	Computational Linguistics in the Netherlands Journal
Volume	10
Status	Published - 13 dec 2020
Evenement	COMPUTATIONAL LINGUISTICS IN THE NETHERLANDS - Utrecht University, Utrecht, Netherlands Duur: 30 jan 2020 → … Congresnummer: 30

Bibliografische nota

Funding Information:
Part of this research was done as an internship at the Instituut voor de Nederlandse Taal by Silke Creten, who is currently working as a PhD student at KU Leuven. Peter Dekker was supported by funding from the Flemish Government under the Onderzoeksprogramma Artificiële Intelligentie (AI) Vlaanderen. programme. Supervision of the project was done by Vincent Vandeghinste, senior researcher at the Instituut voor de Nederlandse Taal and part-time researcher at Leuven.AI and the Centre for Computational Linguistics of KU Leuven. We want to thank the anonymous reviewers for their valuable suggestions that have helped greatly improve the present article.

Funding Information:
Part of this research was done as an internship at the Instituut voor de Nederlandse Taal by Silke Creten, who is currently working as a PhD student at KU Leuven. Peter Dekker was supported by funding from the Flemish Government under the Onderzoeksprogramma Artificiële Intelligentie (AI) Vlaanderen. programme. Supervision of the project was done by Vincent Vandeghinste, senior researcher at the Instituut voor de Nederlandse Taal and part-time researcher at Leuven.AI and the Centre for Computational Linguistics of KU Leuven.

Publisher Copyright:
© 2020 Silke Creten, Peter Dekker, Vincent Vandeghinste.

Copyright:
Copyright 2021 Elsevier B.V., All rights reserved.

Toegang tot document

LInguistic enrichtment of historical Dutch using deep learningFinal published version, 261 KBLicentie: CC BY

VLAAI1: Subsidie: Onderzoeksprogramma Artificiële Intelligentie (AI) Vlaanderen
Nowe, A. & Vanderborght, B.
1/07/19 → 31/12/23
Project: Toegepast

Linguistic enrichment of historical Dutch using deep learning
Silke Creten (Speaker), Peter Dekker (Speaker) & Vincent Vandeghinste (Speaker)
30 jan 2020
Activiteit: Talk or presentation at a conference

Citeer dit

@article{552ccfd61b3f4ef4b660dde919b47996,

title = "Linguistic Enrichment of Historical Dutch using Deep Learning",

abstract = "This article discusses the automatic linguistic enrichment of historical Dutch corpora through the use of part-of-speech tagging and lemmatization. Such a type of enrichment facilitates linguistic research where manual annotation is unfeasible. We built a neural network-based model using the PIE framework and performed an in-depth error analysis, in order to identify the strengths and weaknesses of each approach with respect to labeling historical data. In order to do so, we experimented with two data sets: the Corpus Gysseling (13th century texts) and the Corpus van Reenen/Mulder (14th century texts). We used two different statistical approaches (MBT and HunPos) as baselines for our neural approach. MBT is a memory-based tagger frequently used for modern Dutch, while HunPos is an open source trigram tagger. We present thoroughly analyzed results. In general, the neural model scores better than the two baselines, even with limited training data. Based on the error analysis, we propose several strategies for future research in order to improve the labeling of historical Dutch.",

keywords = "linguistic enrichment, pos tagging, neural networks, deep learning, historical linguistics, dutch, historical dutch",

author = "Silke Creten and Peter Dekker and Vincent Vandeghinste",

note = "Funding Information: Part of this research was done as an internship at the Instituut voor de Nederlandse Taal by Silke Creten, who is currently working as a PhD student at KU Leuven. Peter Dekker was supported by funding from the Flemish Government under the Onderzoeksprogramma Artifici{\"e}le Intelligentie (AI) Vlaanderen. programme. Supervision of the project was done by Vincent Vandeghinste, senior researcher at the Instituut voor de Nederlandse Taal and part-time researcher at Leuven.AI and the Centre for Computational Linguistics of KU Leuven. We want to thank the anonymous reviewers for their valuable suggestions that have helped greatly improve the present article. Funding Information: Part of this research was done as an internship at the Instituut voor de Nederlandse Taal by Silke Creten, who is currently working as a PhD student at KU Leuven. Peter Dekker was supported by funding from the Flemish Government under the Onderzoeksprogramma Artifici{\"e}le Intelligentie (AI) Vlaanderen. programme. Supervision of the project was done by Vincent Vandeghinste, senior researcher at the Instituut voor de Nederlandse Taal and part-time researcher at Leuven.AI and the Centre for Computational Linguistics of KU Leuven. Publisher Copyright: {\textcopyright} 2020 Silke Creten, Peter Dekker, Vincent Vandeghinste. Copyright: Copyright 2021 Elsevier B.V., All rights reserved.; COMPUTATIONAL LINGUISTICS IN THE NETHERLANDS ; Conference date: 30-01-2020",

year = "2020",

month = dec,

day = "13",

language = "English",

volume = "10",

pages = "57--72",

journal = "Computational Linguistics in the Netherlands",

issn = "2211-4009",

}

TY - JOUR

T1 - Linguistic Enrichment of Historical Dutch using Deep Learning

AU - Creten, Silke

AU - Dekker, Peter

AU - Vandeghinste, Vincent

N1 - Conference code: 30

PY - 2020/12/13

Y1 - 2020/12/13

N2 - This article discusses the automatic linguistic enrichment of historical Dutch corpora through the use of part-of-speech tagging and lemmatization. Such a type of enrichment facilitates linguistic research where manual annotation is unfeasible. We built a neural network-based model using the PIE framework and performed an in-depth error analysis, in order to identify the strengths and weaknesses of each approach with respect to labeling historical data. In order to do so, we experimented with two data sets: the Corpus Gysseling (13th century texts) and the Corpus van Reenen/Mulder (14th century texts). We used two different statistical approaches (MBT and HunPos) as baselines for our neural approach. MBT is a memory-based tagger frequently used for modern Dutch, while HunPos is an open source trigram tagger. We present thoroughly analyzed results. In general, the neural model scores better than the two baselines, even with limited training data. Based on the error analysis, we propose several strategies for future research in order to improve the labeling of historical Dutch.

AB - This article discusses the automatic linguistic enrichment of historical Dutch corpora through the use of part-of-speech tagging and lemmatization. Such a type of enrichment facilitates linguistic research where manual annotation is unfeasible. We built a neural network-based model using the PIE framework and performed an in-depth error analysis, in order to identify the strengths and weaknesses of each approach with respect to labeling historical data. In order to do so, we experimented with two data sets: the Corpus Gysseling (13th century texts) and the Corpus van Reenen/Mulder (14th century texts). We used two different statistical approaches (MBT and HunPos) as baselines for our neural approach. MBT is a memory-based tagger frequently used for modern Dutch, while HunPos is an open source trigram tagger. We present thoroughly analyzed results. In general, the neural model scores better than the two baselines, even with limited training data. Based on the error analysis, we propose several strategies for future research in order to improve the labeling of historical Dutch.

KW - linguistic enrichment

KW - pos tagging

KW - neural networks

KW - deep learning

KW - historical linguistics

KW - dutch

KW - historical dutch

UR - http://www.scopus.com/inward/record.url?scp=85120974786&partnerID=8YFLogxK

M3 - Article

VL - 10

SP - 57

EP - 72

JO - Computational Linguistics in the Netherlands

JF - Computational Linguistics in the Netherlands

SN - 2211-4009

T2 - COMPUTATIONAL LINGUISTICS IN THE NETHERLANDS

Y2 - 30 January 2020

ER -

Linguistic Enrichment of Historical Dutch using Deep Learning

Samenvatting

Bibliografische nota

Toegang tot document

Vingerafdruk

Projecten

VLAAI1: Subsidie: Onderzoeksprogramma Artificiële Intelligentie (AI) Vlaanderen

Activiteiten

Linguistic enrichment of historical Dutch using deep learning

Citeer dit