Overview
When people communicate, they rely on a large body of shared common sense knowledge in order to understand each other. Many barriers we face today in artificial intelligence and user interface design are due to the fact that computers do not share this knowledge. To improve computers' understanding of the world that people live in and talk about, we need to provide them with usable knowledge about the basic relationships between things that nearly every person knows.
In 1999, we began a project at the MIT Media Lab to collect common sense from volunteers on the internet. Nearly ten years later our project has expanded to encompass many different areas, languages, and problems. Currently, the English site has over 750,000 sentences from over 15,000 contributers. There are OMCSes in Korean, Japanese (temporarily down), Portuguese, and soon Dutch.
Our research addresses:
- Creating systems that understand the connections between everyday events and objects, people's beliefs, an the way they express them in language,
- Using this understanding to make computers more "people-friendly"
- Developing representations for different varieties of common sense knowledge
- Developing methods for acquiring common sense knowledge from people, corpora, and the web
- Developing architectures that let us fuse these diverse techniques into flexible and resourceful systems
Want to use our software in your own project? Getting Started
