论文标题
语言模型了解测量吗？
Do Language Models Understand Measurements?
论文作者
论文摘要
预训练的语言模型（PLM）的最新成功激发了对他们理解和使用数字的能力的兴趣。然而，尽管有重要性，但尚未对测量的数值推理进行正式研究。在这项研究中，我们表明PLM缺乏对测量值进行推理所需的能力。此外，我们发现在测量丰富的语料库上训练的语言模型在理解测量方面表现出更好的表现。我们提出了一种简单的嵌入策略，以更好地区分数字和单位，从而导致探测任务的显着改善。
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.
