Dan G. Tecuci

My dissertation research investigates generic ways for an intelligent agent to organize and retrieve its past experience in an accurate and scalable way.

I have developed a generic episodic memory module that can be attached to a variety of applications and used for different tasks. Encapsulating the complexity of such a memory into a separate subsystem should reduce the complexity the overall system while allowing research to focus on the generic aspects of memory organization and retrieval in isolation of a specific domain and task.

As part of the Knowledge Systems Research Group I've worked on various knowledge representation and reasoning topics:

• developing contents for our Component Library (a library of generic, composable concepts).
  See these papers (AAAI-97 and K-Cap-01) for more details.
• providing semantic-based access to these components (the semantic search facility in the Component Library Browser).
• developing knowledge-based systems capable of answering novel AP-level questions (and explain their answers) in a broad range of scientific disciplines (like Chemistry, Physics and Biology).
  See our AI Magazine article and our KR-04 paper.
• learning new knowledge from text, by combining NLP and KR&R techniques.
  See our AAAI-07 paper.
• investigating the source of brittleness in answering questions posed in natural language.
  See our K-Cap-07 paper.

I have also done some work in spatial reasoning and cognitive modelling. More specifically, we tested the "Skeleton in the Cognitive Map" hypothesis (read more about this) that states that experts find routes in complex environments by finding a connection from the starting place to a subset of major paths (aka `skeleton'), then move within the skeleton close to the destination, and then make a final connection to the destination. This was joint work with Ben Kuipers and Brian Stankiewicz.

This journal paper describes our main result: an algorithm that accurately models human way-finding in complex environments. More details about the algorithm can be found in this research note. Better results on a larger dataset are reported in this presentation.