Parsers know best

• Verfasst von: Do, Bich-Ngoc [VerfasserIn]
• Verfasst von: Rehbein, Ines [VerfasserIn]
• Titel: Parsers know best
• Titelzusatz: German PP attachment revisited
• Verf.angabe: Bich-Ngoc Do, Ines Rehbein
• Jahr: 2020
• Umfang: 13 S.
• Fussnoten: Gesehen am 29.11.2023
• Titel Quelle: Enthalten in: International Conference on Computational Linguistics (28. : 2020 : Online)The 28th International Conference on Computational Linguistics - proceedings of the conference
• Ort Quelle: [Praha, Czech Republic] : International Committee on Computational Linguistics, 2020
• Jahr Quelle: 2020
• Band/Heft Quelle: (2020), Seite 2049-2061
• ISBN Quelle: 978-1-952148-27-9
• DOI: doi:10.18653/v1/2020.coling-main.185
• Datenträger: Online-Ressource
• Sprache: eng
• Bibliogr. Hinweis: Forschungsdaten: Do, Bich-Ngoc, 1989 - : Tool for extracting PP attachment disambiguation dataset
• Bibliogr. Hinweis: Forschungsdaten: Do, Bich-Ngoc, 1989 - : Neural PP attachment disambiguation systems
• Bibliogr. Hinweis: Forschungsdaten: Do, Bich-Ngoc, 1989 - : Real-world PP attachment disambiguation dataset
• Bibliogr. Hinweis: Forschungsdaten: Do, Bich-Ngoc, 1989 - : Neural dependency parser with biaffine attention and BERT embeddings
• Bibliogr. Hinweis: Forschungsdaten: Do, Bich-Ngoc, 1989 - : Topological field labeler for German
• K10plus-PPN: 1871576784

Abstract

In the paper, we revisit the PP attachment problem which has been identified as one of the major sources for parser errors and discuss shortcomings of recent work. In particular, we show that using gold information for the extraction of attachment candidates as well as a missing comparison of the system's output to the output of a full syntactic parser leads to an overly optimistic assessment of the results. We address these issues by presenting a realistic evaluation of the potential of different PP attachment systems, using fully predicted information as system input. We compare our results against the output of a strong neural parser and show that the full parsing approach is superior to modeling PP attachment disambiguation as a separate task.