When we listen to a podcast or read a text, our brains process the words not by how they are defined individually, but by the concepts and relationships that they evoke in context. Most words are best understood on the basis of a semantic frame, a description of a type of event, relation, or entity and the participants in it. This suggests that if we want computers to be able to understand language the way we do, we need to teach them to understand semantic frames.
For over 25 years, ICSI researchers have been doing just that. With support from the National Science Foundation, the Defense Threat Reduction Agency, and other funders, Charles J. Fillmore, Collin Baker, and colleagues developed FrameNet, a lexical database that defines more than 13,000 lexical units (word senses) in over 1,000 semantic frames spanning domains from mining to morality. The system draws from over 200,000 manual annotations documenting how words are used in actual texts and is designed to be readable by both machines and people.
Teaching computers not just words, but meaning
FrameNet’s rich trove of data, freely available for download, has proved valuable for research in machine learning, information extraction, and reasoning about events, as well as for deepening human understanding of vocabulary and grammar. The team is planning for a major expansion of FrameNet that will combine linguistic expertise, crowdsourced data, and machine learning to improve both computers’ understanding of natural language and human awareness of exactly what computers understand. In addition, visitors from many countries have learned about Frame Semantics at ICSI and returned to their home institutions to found FrameNet-like projects in eight additional languages.
What made ICSI a good place to pursue these projects?
“FrameNet would not have existed without ICSI. My work at ICSI has been my career for more than 25 years. ICSI has consistently been supportive of our research and the FrameNet project in general, and I would recommend ICSI to others because of its history of supporting a wide range of basic research, both intellectually and administratively.
FrameNet has also been fortunate to have welcomed many DAAD visitors over the years, and the project has greatly benefitted from their presence. The questions they have asked have repeatedly helped us think more deeply about Frame Semantic theory. Most of them published FrameNet-related papers while at ICSI and now teach at the university level.”
Collin F. Baker
Senior Fellow, ICSI
This story was published in January 2026 as part of a retrospective series highlighting ICSI’s accomplishments and impacts over the years. To learn about our ongoing work, explore our Core Research Themes.