Do Language Models Understand Measurements?

Autor:	Park, Sungjin, Ryu, Seungwoo, Choi, Edward
Rok vydání:	2022
Předmět:	Computer Science - Computation and Language
Druh dokumentu:	Working Paper
Popis:	Recent success of pre-trained language models (PLMs) has stimulated interest in their ability to understand and work with numbers. Yet, the numerical reasoning over measurements has not been formally studied despite their importance. In this study, we show that PLMs lack the capability required for reasoning over measurements. Furthermore, we find that a language model trained on a measurement-rich corpus shows better performance on understanding measurements. We propose a simple embedding strategy to better distinguish between numbers and units, which leads to a significant improvement in the probing tasks. Comment: Findings of EMNLP 2022
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2210.12694 Zobrazit plný text záznamu View this record from Arxiv