Work in the Media Lab

ROADIE (Applying Commonsense Reasoning to Consumer Electronic Interaction): Consumer electronics devices are becoming more and more complicated to the point that the user is scare to manipulate them. These devices do not know anything about every day life and human goals and they show irrelevant menus and options. By using EventNet, a comonsense reasoning plan recognizer, we build an interface with knowledge about the user’s intentions. This knowledge helps the device to display relevant information to reach the user’s goal. For example, a living room set that knows how to configure itself to watch the news. This leads to a more human like interaction with these devices.
Download master thesis (pdf)
Video (720x480 128MB)
Video (320x240 56MB)

EventNet: EventNet is the commonsense approach to plan recognizing. By extracting cause-effect and action-goal links from Open Mind Common Sense, we build an associative network of tens of thousands of nodes about human actions and goals. For example, EventNet is able to infer that in order to go from home to the movies you should drive your car. It is a toolkit for the easy use in interactive applications.

Using Commonsense Reasoning to Find Cultural Differences in Text: Because commonsense knowledge differs in each culture, misunderstandings frequently occur. Since differences can be subtle, there has been little work in trying to detect places in text where cultural differences might arise. We explicitly represent commonsense knowledge of each culture in separate knowledge bases. By analyzing the text, we can find the differences between each culture’s knowledge concerning the subject of the text.

GloBuddy 2: When traveling in foreign countries, people often rely on traditional phrase books for language translation.  However, these phrase books only work in a limited number of common situations, and even common situations will often deviate from the predefined script the phrase book relies on.  Using a vast knowledge base of commonsense facts and relationships, GloBuddy 2 is able to expand on the user’s translation request and provide words and phrases related to the user’s situation.

Using Commonsense Reasoning in Video Game Design: When computers automatically generate video game environments, the result is often as simplistic as a randomly generated maze. Using the Open Mind Common Sense knowledge base, we are exploring ways to automatically generate virtual environments that are immersive and intelligently designed. By allowing video games to dynamically create their own environments, developers will be able to reduce the time they currently spend crafting environments by hand and focus on higher level design issues. More importantly, games that are capable of changing their own environments will feel less static, resulting in a more believable experience and an increased replay value.

Using Commonsense Reasoning to Improve Voice Recognition: Current voice recognition software relies on statistical techniques to determine which words a user has said. In this project we are attempting to leverage the semantic context of what the user has previously said to improve future predictions. We are using ConceptNet, a semantic network created from the Open Mind Common Sense knowledge base to disambiguate phonetically similar words and improve overall recognition accuracy.

Work before coming to the media lab