Collaborating Authors


Language Models are Open Knowledge Graphs Artificial Intelligence

This paper shows how to construct knowledge graphs (KGs) from pre-trained language models (e.g., BERT, GPT-2/3), without human supervision. Popular KGs (e.g, Wikidata, NELL) are built in either a supervised or semi-supervised manner, requiring humans to create knowledge. Recent deep language models automatically acquire knowledge from large-scale corpora via pre-training. The stored knowledge has enabled the language models to improve downstream NLP tasks, e.g., answering questions, and writing code and articles. In this paper, we propose an unsupervised method to cast the knowledge contained within language models into KGs. We show that KGs are constructed with a single forward pass of the pre-trained language models (without fine-tuning) over the corpora. We demonstrate the quality of the constructed KGs by comparing to two KGs (Wikidata, TAC KBP) created by humans. Our KGs also provide open factual knowledge that is new in the existing KGs. Our code and KGs will be made publicly available.

Thomas Cover, acclaimed information theorist and electrical engineer, dies at 73

AITopics Original Links

Cover was a man of remarkable breadth in his research interests, making landmark contributions in fields ranging from information theory and mathematical statistics to data compression, pattern recognition and stock market investment strategies. Thomas Cover, one of the world's top information theorists and a professor of electrical engineering and of statistics at Stanford University, died March 26 at Stanford Hospital at the age of 73. "A senior colleague at MIT often referred to Tom Cover as'the jewel in Stanford's crown.' He certainly was," said Abbas El Gamal, professor of electrical engineering. "Not only was he one of the greatest information theorists since Claude Shannon (and the one most like Shannon in approach, clarity, and taste), but he was also one of Stanford's most inspiring teachers and mentors.