Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2204.08831 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914805059682304 |
|---|---|
| author | Lasri, Karim Pimentel, Tiago Lenci, Alessandro Poibeau, Thierry Cotterell, Ryan |
| author_facet | Lasri, Karim Pimentel, Tiago Lenci, Alessandro Poibeau, Thierry Cotterell, Ryan |
| contents | A central quest of probing is to uncover how pre-trained models encode a linguistic property within their representations. An encoding, however, might be spurious-i.e., the model might not rely on it when making predictions. In this paper, we try to find encodings that the model actually uses, introducing a usage-based probing setup. We first choose a behavioral task which cannot be solved without using the linguistic property. Then, we attempt to remove the property by intervening on the model's representations. We contend that, if an encoding is used by the model, its removal should harm the performance on the chosen behavioral task. As a case study, we focus on how BERT encodes grammatical number, and on how it uses this encoding to solve the number agreement task. Experimentally, we find that BERT relies on a linear encoding of grammatical number to produce the correct behavioral output. We also find that BERT uses a separate encoding of grammatical number for nouns and verbs. Finally, we identify in which layers information about grammatical number is transferred from a noun to its head verb. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2204_08831 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | Probing for the Usage of Grammatical Number Lasri, Karim Pimentel, Tiago Lenci, Alessandro Poibeau, Thierry Cotterell, Ryan Computation and Language A central quest of probing is to uncover how pre-trained models encode a linguistic property within their representations. An encoding, however, might be spurious-i.e., the model might not rely on it when making predictions. In this paper, we try to find encodings that the model actually uses, introducing a usage-based probing setup. We first choose a behavioral task which cannot be solved without using the linguistic property. Then, we attempt to remove the property by intervening on the model's representations. We contend that, if an encoding is used by the model, its removal should harm the performance on the chosen behavioral task. As a case study, we focus on how BERT encodes grammatical number, and on how it uses this encoding to solve the number agreement task. Experimentally, we find that BERT relies on a linear encoding of grammatical number to produce the correct behavioral output. We also find that BERT uses a separate encoding of grammatical number for nouns and verbs. Finally, we identify in which layers information about grammatical number is transferred from a noun to its head verb. |
| title | Probing for the Usage of Grammatical Number |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2204.08831 |