Explaining pretrained language models' understanding of linguistic structures using construction grammar
Construction Grammar (CxG) is a paradigm from cognitive linguistics emphasizing the connection between syntax and semantics. Rather than rules that operate on lexical items, it posits constructions as the central building blocks of language, i.e., linguistic units of different granularity that combi...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Frontiers Media S.A.
2023-10-01
|
Series: | Frontiers in Artificial Intelligence |
Subjects: | |
Online Access: | https://www.frontiersin.org/articles/10.3389/frai.2023.1225791/full |
_version_ | 1797661268704755712 |
---|---|
author | Leonie Weissweiler Leonie Weissweiler Valentin Hofmann Valentin Hofmann Abdullatif Köksal Abdullatif Köksal Hinrich Schütze Hinrich Schütze |
author_facet | Leonie Weissweiler Leonie Weissweiler Valentin Hofmann Valentin Hofmann Abdullatif Köksal Abdullatif Köksal Hinrich Schütze Hinrich Schütze |
author_sort | Leonie Weissweiler |
collection | DOAJ |
description | Construction Grammar (CxG) is a paradigm from cognitive linguistics emphasizing the connection between syntax and semantics. Rather than rules that operate on lexical items, it posits constructions as the central building blocks of language, i.e., linguistic units of different granularity that combine syntax and semantics. As a first step toward assessing the compatibility of CxG with the syntactic and semantic knowledge demonstrated by state-of-the-art pretrained language models (PLMs), we present an investigation of their capability to classify and understand one of the most commonly studied constructions, the English comparative correlative (CC). We conduct experiments examining the classification accuracy of a syntactic probe on the one hand and the models' behavior in a semantic application task on the other, with BERT, RoBERTa, and DeBERTa as the example PLMs. Our results show that all three investigated PLMs, as well as OPT, are able to recognize the structure of the CC but fail to use its meaning. While human-like performance of PLMs on many NLP tasks has been alleged, this indicates that PLMs still suffer from substantial shortcomings in central domains of linguistic knowledge. |
first_indexed | 2024-03-11T18:42:58Z |
format | Article |
id | doaj.art-9d663f4504a94e04b009cd0876bef3ee |
institution | Directory Open Access Journal |
issn | 2624-8212 |
language | English |
last_indexed | 2024-03-11T18:42:58Z |
publishDate | 2023-10-01 |
publisher | Frontiers Media S.A. |
record_format | Article |
series | Frontiers in Artificial Intelligence |
spelling | doaj.art-9d663f4504a94e04b009cd0876bef3ee2023-10-12T09:06:41ZengFrontiers Media S.A.Frontiers in Artificial Intelligence2624-82122023-10-01610.3389/frai.2023.12257911225791Explaining pretrained language models' understanding of linguistic structures using construction grammarLeonie Weissweiler0Leonie Weissweiler1Valentin Hofmann2Valentin Hofmann3Abdullatif Köksal4Abdullatif Köksal5Hinrich Schütze6Hinrich Schütze7Center for Information and Language Processing, LMU Munich, Munich, GermanyMunich Center for Machine Learning, Munich, GermanyCenter for Information and Language Processing, LMU Munich, Munich, GermanyFaculty of Linguistics, University of Oxford, Oxford, United KingdomCenter for Information and Language Processing, LMU Munich, Munich, GermanyMunich Center for Machine Learning, Munich, GermanyCenter for Information and Language Processing, LMU Munich, Munich, GermanyMunich Center for Machine Learning, Munich, GermanyConstruction Grammar (CxG) is a paradigm from cognitive linguistics emphasizing the connection between syntax and semantics. Rather than rules that operate on lexical items, it posits constructions as the central building blocks of language, i.e., linguistic units of different granularity that combine syntax and semantics. As a first step toward assessing the compatibility of CxG with the syntactic and semantic knowledge demonstrated by state-of-the-art pretrained language models (PLMs), we present an investigation of their capability to classify and understand one of the most commonly studied constructions, the English comparative correlative (CC). We conduct experiments examining the classification accuracy of a syntactic probe on the one hand and the models' behavior in a semantic application task on the other, with BERT, RoBERTa, and DeBERTa as the example PLMs. Our results show that all three investigated PLMs, as well as OPT, are able to recognize the structure of the CC but fail to use its meaning. While human-like performance of PLMs on many NLP tasks has been alleged, this indicates that PLMs still suffer from substantial shortcomings in central domains of linguistic knowledge.https://www.frontiersin.org/articles/10.3389/frai.2023.1225791/fullNLPprobingconstruction grammarcomputational linguisticslarge language models |
spellingShingle | Leonie Weissweiler Leonie Weissweiler Valentin Hofmann Valentin Hofmann Abdullatif Köksal Abdullatif Köksal Hinrich Schütze Hinrich Schütze Explaining pretrained language models' understanding of linguistic structures using construction grammar Frontiers in Artificial Intelligence NLP probing construction grammar computational linguistics large language models |
title | Explaining pretrained language models' understanding of linguistic structures using construction grammar |
title_full | Explaining pretrained language models' understanding of linguistic structures using construction grammar |
title_fullStr | Explaining pretrained language models' understanding of linguistic structures using construction grammar |
title_full_unstemmed | Explaining pretrained language models' understanding of linguistic structures using construction grammar |
title_short | Explaining pretrained language models' understanding of linguistic structures using construction grammar |
title_sort | explaining pretrained language models understanding of linguistic structures using construction grammar |
topic | NLP probing construction grammar computational linguistics large language models |
url | https://www.frontiersin.org/articles/10.3389/frai.2023.1225791/full |
work_keys_str_mv | AT leonieweissweiler explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar AT leonieweissweiler explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar AT valentinhofmann explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar AT valentinhofmann explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar AT abdullatifkoksal explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar AT abdullatifkoksal explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar AT hinrichschutze explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar AT hinrichschutze explainingpretrainedlanguagemodelsunderstandingoflinguisticstructuresusingconstructiongrammar |