Do Language Models Know the Way to Rome?
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Standard
Do Language Models Know the Way to Rome? / Liétard, Bastien Nathan; Abdou, Mostafa ; Søgaard, Anders.
Proceedings of the Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP. Association for Computational Linguistics, 2021. p. 510–517.Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Do Language Models Know the Way to Rome?
AU - Liétard, Bastien Nathan
AU - Abdou, Mostafa
AU - Søgaard, Anders
PY - 2021
Y1 - 2021
N2 - The global geometry of language models is important for a range of applications, but language model probes tend to evaluate rather local relations, for which ground truths are easily obtained. In this paper we exploit the fact that in geography, ground truths are available beyond local relations. In a series of experiments, we evaluate the extent to which language model representations of city and country names are isomorphic to real-world geography, e.g., if you tell a language model where Paris and Berlin are, does it know the way to Rome? We find that language models generally encode limited geographic information, but with larger models performing the best, suggesting that geographic knowledge can be induced from higher-order co-occurrence statistics.
AB - The global geometry of language models is important for a range of applications, but language model probes tend to evaluate rather local relations, for which ground truths are easily obtained. In this paper we exploit the fact that in geography, ground truths are available beyond local relations. In a series of experiments, we evaluate the extent to which language model representations of city and country names are isomorphic to real-world geography, e.g., if you tell a language model where Paris and Berlin are, does it know the way to Rome? We find that language models generally encode limited geographic information, but with larger models performing the best, suggesting that geographic knowledge can be induced from higher-order co-occurrence statistics.
U2 - 10.18653/v1/2021.blackboxnlp-1.40
DO - 10.18653/v1/2021.blackboxnlp-1.40
M3 - Article in proceedings
SP - 510
EP - 517
BT - Proceedings of the Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP
PB - Association for Computational Linguistics
T2 - Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP
Y2 - 11 November 2021 through 11 November 2021
ER -
ID: 300078921