MIT Media Lab Course - Fall 2006 Common Sense Reasoning for Interactive Applications
Henry Lieberman |
Student Projects Click on the title/picture for the paper on the project |
Common Sense Video Game
Authoring
Jeff Orkin
In this paper we describe how integrating a common sense knowledge base with a commercial video game engine allows us to automate AI character behavior. Today’s video game characters can animate beautifully and navigate intelligently from place to place, but they require detailed instructions at a low level about where to go and how to interact with the environment. Associating pre-existing human-labeled game assets with common sense knowledge provides AI characters with an understanding of object affordances, and goals of people in various social roles. We illustrate how this integration leads to automation by describing an implementation of a proof-of-concept common sense game authoring system used to create a restaurant game, where a waitress serves a hungry patron.
Storied Navigation Using Common Sense
Hector Beltran and Edward Shen
Human beings love to tell stories. Much of our daily interaction revolves around exchanging stories. Using modern technologies, tools are readily-available that allow us to share our stories with media such as photos or videos. This paper presents the first steps to creating an intelligent system that would allow users to interactively tell stories, namely, Storied Navigation. By using commonsense reasoning technology and analogies to stories that have been previously entered by other users, the system shows potential to deliver an enjoyable storytelling experience.
AffectiveSynthesizer:
An Affective Text to Speech System
Emotions have huge impact on people’s voice. Research projects, which apply emotion impacts to speech synthesis, start to emerge recently. This paper described how common sense reasoning technologies could help text to speech (TTS) system understand emotions behind input texts, so the arguments of TTS system can be fine-tuned accordingly to synthesize speech correspondent to the emotions reside within the input texts, which opens the door for using synthesized speech as the production tool of audio applications/services.
CSAMOA:
A Common Sense Application Model of
Architecture
Jason B. Alonso
It has become apparent that many human-computer interface
applications of common sense reasoning, particularly those
built on theOpenMind Common Sense corpus , make use of
similar computational tools (spreading activation, for example)
in addition to the corpus itself. Meanwhile, new representations,
new methods of reasoning, and new applications
are being introduced without a clear foundation for understanding
their interrelationships. In this paper, I describe my
goals and progress in the design of a model of architecture
for expressing the interoperation of common sense tools developed
as parts of different efforts.
CSAMOA and ConceptNet: An
experimental application of
a model of architecture
Jason B. Alonso
The Common Sense Application Model of Architecture
(CSAMOA) was created to organize applications, libraries,
and corpora of common sense reasoning with a concise taxonomy
and guidelines that encurage sustainable code. In this
paper, I discuss the motivations for selecting ConceptNet for
reimplementation, compromises made in the enforcement of
the architecture, observations on the effectiveness of the architecture,
the future of this experiment, and plans for additional
experiments.
Acquisition of Procedural Commonsense
Knowledge in a
3-D Simulated World
Moin Ahmad
In this paper we describe the design of a solution for the
acquisition of procedural commonsense knowledge in a
situated simulated environment. In this system, users in a
simulated game environment, control virtual robots to do
various tasks in a real world scenario. In the current system,
users are situated in a restaurant scenario and take charge of
teaching their apprentice robots actions to complete tasks.
The main benefit of situating acquisition of knowledge in a
real world environment is that procedures and actions in
situations get learnt in many different realms of thought and
at many different levels of detail ultimately learning a task
in many ways.