We aim at better understanding scientific papers and information extraction. Examples include structured tuples (e.g. entities and relations), text snippets (e.g. claims), and other paper objects (e.g. figures, tables, equations). We use these to automatically construct knowledge bases and facilitate...
In this project, we investigate language modeling approaches for scientific documents. Our goal is to provide general language models (like BERT) or other approaches that could be used for many tasks relevant to the scientific domain.
The increasing growth of scientific literature has made discovering new and relevant content and keeping up with the latest developments in the field challenging. In this project, we investigate methods to best discover and present relevant literature to users.
Ontologies are used to ground lexical items in various NLP tasks including entity linking, question answering, semantic parsing and information retrieval. In this project we investigate approaches for improving and constructing domain-specific ontologies using NLP methods.
Generating natural language content relevant to scientific domain is a challenging task. In this project we investigate approaches for text generation and summarization of scientific papers.
To solve the hardest problems, AI systems must work together with humans. We design and study interactive AI systems that improve outcomes compared to what is possible with AI or people alone.
Using the Semantic Scholar corpus of academic publications, we perform large-scale analyses of the scientific literature - discovering trends, identifying biases, and illuminating existing structures.