Guest Talk "Domain Adaptive Neurosymbolic Agents: KASTLE's efforts towards DIANA"

19/08/2025

Cogan Shimizu 

Date/Time: 26.08.2025, 12:00 

Location: D2.2.094 

Abstract 

The ability for AI agents to generalize and adapt across domains is paramount. We (KASTLE Lab) are tackling this (re-)emerging frontier through a proposed framework, which we call: *Domain Adaptive NeuroSymbolic Agents (DIANA)*. This talk invites discussion on the various components that are currently under investigation: (a) pattern-based methods for domain adaptation from unstructured natural language or documentation; (b) neurosemantic approaches for knowledge gap discovery and designing the steps to fill them; and (c) a neurosymbolic mechanism for evolving knowledge stemming from those activities. 

Bio 

Cogan Shimizu is an Assistant Professor at Wright State University in the Department of Computer Science & Engineering, where he direct the Knowledge and Semantic Technologies (KASTLE) Lab. He was a member the Data Semantics Laboratory (Kansas State University) and completed my M.S. and Ph.D. under the supervision of Dr. Pascal Hitzler. His research spans the entire spectrum of the ontology development pipeline, essentially examining first the pre-conditions for constructing well-founded ontologies all the way through their use in downstream tasks, such as embeddings or machine learning.

Back to overview