It is shown that materials science knowledge present in the published literature can be efficiently encoded as information-dense word embeddings11–13 (vector representations of words) without human labelling or supervision, suggesting that latent knowledge regarding future discoveries is to a large extent embedded in past publications.
使用 AI 将内容摘要翻译为中文，便于快速阅读
使用 AI 分析这篇文章的核心发现、关键要点和深度见解
由 DeepSeek AI 提供分析 · 首次使用需配置 API Key
arXiv · 2006-03-27
arXiv · 2024-03-06
arXiv · 2021-08-10
arXiv · 2018-07-03
arXiv · 2019-03-13