Journal Article

Measuring and Improving Consistency in Pretrained Language Models

Schütze, Hinrich, Elazar, Yanai, Goldberg, Yoav, Kassner, Nora, Ravichander, Abhilasha, Hovy, Eduard, Ravfogel, Shauli

2021

Overview

of a model—that is, the invariance of its behavior under meaning-preserving alternations in its input—is a highly desirable property in natural language processing. In this paper we study the question: Are Pretrained Language Models (PLMs) consistent with respect to factual knowledge? To this end, we create 🤘, a high-quality resource of cloze-style query English paraphrases. It contains a total of 328 paraphrases for 38 relations. Using 🤘, we show that the consistency of all PLMs we experiment with is poor— though with high variance between relations. Our analysis of the representational spaces of PLMs suggests that they have a poor structure and are currently not suitable for representing knowledge robustly. Finally, we propose a method for improving model consistency and experimentally demonstrate its effectiveness.

• Publisher: MIT Press,MIT Press Journals, The,The MIT Press
• Subject: Accuracy / Alternations / Consistency / Knowledge / Knowledge representation / Language / Language modeling / Linguistics / Natural language / Natural language processing / Paraphrase / Queries / Variance analysis