Using Integrated Gradients and Constituency Parse Trees to explain Linguistic Acceptability learnt by BERT

Anmol Nayak, Hari Prasad Timmapathini


Abstract
Linguistic Acceptability is the task of determining whether a sentence is grammatical or ungrammatical. It has applications in several use cases like Question-Answering, Natural Language Generation, Neural Machine Translation, where grammatical correctness is crucial. In this paper we aim to understand the decision-making process of BERT (Devlin et al., 2019) in distinguishing between Linguistically Acceptable sentences (LA) and Linguistically Unacceptable sentences (LUA).We leverage Layer Integrated Gradients Attribution Scores (LIG) to explain the Linguistic Acceptability criteria that are learnt by BERT on the Corpus of Linguistic Acceptability (CoLA) (Warstadt et al., 2018) benchmark dataset. Our experiments on 5 categories of sentences lead to the following interesting findings: 1) LIG for LA are significantly smaller in comparison to LUA, 2) There are specific subtrees of the Constituency Parse Tree (CPT) for LA and LUA which contribute larger LIG, 3) Across the different categories of sentences we observed around 88% to 100% of the Correctly classified sentences had positive LIG, indicating a strong positive relationship to the prediction confidence of the model, and 4) Around 43% of the Misclassified sentences had negative LIG, which we believe can become correctly classified sentences if the LIG are parameterized in the loss function of the model.
Anthology ID:
2021.icon-main.11
Volume:
Proceedings of the 18th International Conference on Natural Language Processing (ICON)
Month:
December
Year:
2021
Address:
National Institute of Technology Silchar, Silchar, India
Editors:
Sivaji Bandyopadhyay, Sobha Lalitha Devi, Pushpak Bhattacharyya
Venue:
ICON
SIG:
Publisher:
NLP Association of India (NLPAI)
Note:
Pages:
80–85
Language:
URL:
https://aclanthology.org/2021.icon-main.11
DOI:
Bibkey:
Cite (ACL):
Anmol Nayak and Hari Prasad Timmapathini. 2021. Using Integrated Gradients and Constituency Parse Trees to explain Linguistic Acceptability learnt by BERT. In Proceedings of the 18th International Conference on Natural Language Processing (ICON), pages 80–85, National Institute of Technology Silchar, Silchar, India. NLP Association of India (NLPAI).
Cite (Informal):
Using Integrated Gradients and Constituency Parse Trees to explain Linguistic Acceptability learnt by BERT (Nayak & Timmapathini, ICON 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.icon-main.11.pdf
Data
CoLAGLUE