Are Pretrained Language Models Symbolic Reasoners Over Knowledge?
MCML Authors
Nora Kassner
* Former Member
Hinrich Schütze
Prof. Dr.
Core PI
Abstract
Nora Kassner
* Former Member
Hinrich Schütze
Prof. Dr.
Core PI
Abstract
How can pretrained language models (PLMs) learn factual knowledge from the training set? We investigate the two most important mechanisms: reasoning and memorization. Prior work has attempted to quantify the number of facts PLMs learn, but we present, using synthetic data, the first study that investigates the causal relation between facts present in training and facts learned by the PLM. For reasoning, we show that PLMs seem to learn to apply some symbolic reasoning rules correctly but struggle with others, including two-hop reasoning. Further analysis suggests that even the application of learned reasoning rules is flawed. For memorization, we identify schema conformity (facts systematically supported by other facts) and frequency as key factors for its success.
inproceedings KKS20
CoNLL 2020
24th Conference on Computational Natural Language Learning. Virtual, Nov 19-20, 2020.
Authors
N. Kassner • B. Krojer • H. SchützeLinks
DOIResearch Area
BibTeXKey: KKS20