School of Computing, Informatics, and Decision Systems Engineering
At the core of Chitta Baral’s research is understanding knowledge creation and management. He is working to build intelligent programs, systems and robots that have the capacity to reason and learn.
One of the reasons that there are not more intelligent systems is that acquiring knowledge is a complex process. Baral is taking a different approach to artificial intelligence that mimics how humans learn language.
We learn through examples and feedback. Humans may read a sentence and understand the meaning even if they don’t understand every single word. Other factors must be taken into consideration as well, such as words with multiple meanings or how the context of the situation can carry different significance and weight.
By exploring what knowledge looks like and how we can translate that to a computer, Baral aims to create intelligent systems that can take large amounts of data and think with it.
This research is being applied through several projects. For the Office of Naval Research, he is working on human-robot interaction, developing systems that understand and translate spoken directions into a language that robot understands. With IARPA, he is working on technology that can understand text from various sources, identify sociological factors and analyze conversations for contextual clues such as who is the leader of the group. He is also working on an NSF program developing a system that can analyze and disseminate large volumes of text.
While applicable to a wide range of domains, one of particular interest to Baral is biomedical research. Researchers struggle to keep pace with the volumes of new information being published each year. Using reasoning and natural language text, Baral is working on systems that can analyze the volumes of information and provide efficient access to a knowledge base.
In his work going forward, Baral sees the opportunity to apply this type of large-scale information analysis to patient records, making it easier and quicker for medical professionals to read and understand patient histories.
Baral is also working on future training tools by taking a middle school biology book and creating a system that can read a chapter then answer questions.
Baral has been at ASU since 1999. He is author of the 2003 textbook, Knowledge, Representation, Reasoning and Declarative Problem Solving. More information at http://www.public.asu.edu/~cbaral/ and the BioAI Research Lab, http://www.fulton.asu.edu/~bioai/.