Conditional probing: measuring usable information beyond a baseline

John Hewitt, Kawin Ethayarajh, Percy Liang, Christopher Manning


Abstract
Probing experiments investigate the extent to which neural representations make properties—like part-of-speech—predictable. One suggests that a representation encodes a property if probing that representation produces higher accuracy than probing a baseline representation like non-contextual word embeddings. Instead of using baselines as a point of comparison, we’re interested in measuring information that is contained in the representation but not in the baseline. For example, current methods can detect when a representation is more useful than the word identity (a baseline) for predicting part-of-speech; however, they cannot detect when the representation is predictive of just the aspects of part-of-speech not explainable by the word identity. In this work, we extend a theory of usable information called V-information and propose conditional probing, which explicitly conditions on the information in the baseline. In a case study, we find that after conditioning on non-contextual word embeddings, properties like part-of-speech are accessible at deeper layers of a network than previously thought.
Anthology ID:
2021.emnlp-main.122
Volume:
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2021
Address:
Online and Punta Cana, Dominican Republic
Editors:
Marie-Francine Moens, Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1626–1639
Language:
URL:
https://aclanthology.org/2021.emnlp-main.122
DOI:
10.18653/v1/2021.emnlp-main.122
Bibkey:
Cite (ACL):
John Hewitt, Kawin Ethayarajh, Percy Liang, and Christopher Manning. 2021. Conditional probing: measuring usable information beyond a baseline. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 1626–1639, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
Conditional probing: measuring usable information beyond a baseline (Hewitt et al., EMNLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.emnlp-main.122.pdf
Video:
 https://aclanthology.org/2021.emnlp-main.122.mp4
Code
 john-hewitt/conditional-probing
Data
GLUESSTSST-2