Applying Common Sense Using Dimensionality Reduction

1541-1672/09/$26.00 © 2009 IEEE IEEE INTELLIGENT SYSTEMS Published by the IEEE Computer Society conviction that the floor will hold your weight. When people communicate with each other, they rely on similar background knowledge, which they almost never state explicitly. This follows from the maxim of pragmatics that people avoid stating information that is obvious to the listener.1 In the absence of a learning system as complete as the human brain, automatically acquiring all this frequently unstated knowledge would be difficult. But for an AI system to understand the world that humans live in and talk about, it needs to have this unspoken background knowledge. It needs a source of information about the basic relationships between things that nearly every person knows. In one way or another, this implicit knowledge must be made explicit so that a system can use it computationally. The goal of the Open Mind Common Sense (OMCS) project is to provide intuition to AI systems and applications by giving them access to a broad collection of basic knowledge, along with the computational tools to work with it. This knowledge helps applications understand the way objects relate to each other in the world, people’s goals in their daily lives, and the emotional content of events or situations. Reflecting the way people change thought processes and representations to attack different problems, we designed the OMCS system to easily transition between several data formats, using the best representation for an application or problem. Our semantic network, ConceptNet, is built from a corpus of commonsense knowledge collected and rated by volunteers on the Internet. ConceptNet powers our reaWhen we encounter new situations, such as entering an unfamiliar restaurant or store, we rely on our background knowledge to act and