Title: Language convergence in complex, ambiguous situations; and forays Into Security Visualization and Log Anonymization         

Speaker: Kiran Lakkaraju, University of Illinois at Urbana-Champaign

Date/Time: Monday, January 12, 2009, 2:15 – 3:30 pm

Location: CSRI/279

Brief Abstract: Language is a system for encoding and decoding knowledge that allows entities to share information and cooperate to achieve goals. As such, language is a crucial component in any complex, distributed system operating in a dynamic environment. As an example, in the Multi-Agent Systems (MAS) field numerous “Agent Communication Languages” (ACLs) (such as FIPA-ACL or KQML) have been developed to provide agents a means of communicating with each other. In dynamic, complex environments where tasks and goals are rapidly changing language must evolve to represent new learned knowledge. This happens at the individual level. However, language is a collective entity -- it is useful in proportion to the frequency of its use. Thus there are two competing forces, individual language change pushing towards greater variety; and the need for a globally shared language. My work focuses on how to resolve these two competing forces in a manner as to allow for individual variety to arise but also to spread throughout a population. This is called the “Language Convergence Problem”. As an example, consider the current initiative for Net-Centric Operations in the Department of Defense. NCO essentially requires the linking of diverse knowledge sources from a variety of entities, including non-U.S. coalition partners. Effective integration will require semantic mediation between the various partners (consider how the term “full up round” can mean very different things to the Navy and the Army). This is an instance of a language convergence problem, where entities with varying systems of communication need to collaborate and agree upon a shared system.

CSRI POC: Mark D. Rintoul, (505) 844-9592