Using natural language generation to bootstrap missing Wikipedia articles: A human-centric perspective

TY - JOUR

T1 - Using natural language generation to bootstrap missing Wikipedia articles

T2 - A human-centric perspective

AU - Kaffee, Lucie Aimée

AU - Vougiouklis, Pavlos

AU - Simperl, Elena

AU - Cimiano, Philipp

N1 - Funding Information: This research was supported by the EPSRC-funded project Data Stories under grant agreement number EP/P025676/1; and DSTL under grant number DSTLX-1000094186. We would like to express our gratitude to the Wikipedia communities involved in this study for embracing our research and particularly to the participants of the interviews. Hady Elsahar has been an essential part of this work, discussing the ideas with us from the beginning. We thank him for his support and collaboration on this and previous works on the topic of ArticlePlaceholder. Publisher Copyright: © 2022 - The authors. Published by IOS Press.

PY - 2022

Y1 - 2022

N2 - Nowadays natural language generation (NLG) is used in everything from news reporting and chatbots to social media management. Recent advances in machine learning have made it possible to train NLG systems that seek to achieve human-level performance in text writing and summarisation. In this paper, we propose such a system in the context of Wikipedia and evaluate it with Wikipedia readers and editors. Our solution builds upon the ArticlePlaceholder, a tool used in 14 under-resourced Wikipedia language versions, which displays structured data from the Wikidata knowledge base on empty Wikipedia pages. We train a neural network to generate an introductory sentence from the Wikidata triples shown by the ArticlePlaceholder, and explore how Wikipedia users engage with it. The evaluation, which includes an automatic, a judgement-based, and a task-based component, shows that the summary sentences score well in terms of perceived fluency and appropriateness for Wikipedia, and can help editors bootstrap new articles. It also hints at several potential implications of using NLG solutions in Wikipedia at large, including content quality, trust in technology, and algorithmic transparency.

AB - Nowadays natural language generation (NLG) is used in everything from news reporting and chatbots to social media management. Recent advances in machine learning have made it possible to train NLG systems that seek to achieve human-level performance in text writing and summarisation. In this paper, we propose such a system in the context of Wikipedia and evaluate it with Wikipedia readers and editors. Our solution builds upon the ArticlePlaceholder, a tool used in 14 under-resourced Wikipedia language versions, which displays structured data from the Wikidata knowledge base on empty Wikipedia pages. We train a neural network to generate an introductory sentence from the Wikidata triples shown by the ArticlePlaceholder, and explore how Wikipedia users engage with it. The evaluation, which includes an automatic, a judgement-based, and a task-based component, shows that the summary sentences score well in terms of perceived fluency and appropriateness for Wikipedia, and can help editors bootstrap new articles. It also hints at several potential implications of using NLG solutions in Wikipedia at large, including content quality, trust in technology, and algorithmic transparency.

KW - ArticlePlaceholder

KW - Multilingual

KW - Natural language generation

KW - Neural networks

KW - Wikidata

KW - Wikipedia

UR - http://www.scopus.com/inward/record.url?scp=85124722670&partnerID=8YFLogxK

U2 - 10.3233/SW-210431

DO - 10.3233/SW-210431

M3 - Article

AN - SCOPUS:85124722670

SN - 1570-0844

VL - 13

SP - 163

EP - 194

JO - Semantic Web

JF - Semantic Web

IS - 2

ER -