Representing Knowledge with Primitives in Conditional Probabilities

Speaker: Bob Kirby

6:30 – 7:00  Registration, pizza, networking
7:00 – 7:10  Announcements, job openings, book drawing for members
7:10 Talk

Event Details

This research talk espouses using conditional probabilities to represent and process knowledge from natural language, images, and sensed data. Outcomes of a conditional probability assert a statement with quantified uncertainty, assuming the condition. Concepts within the condition and outcomes are high-order logical expressions with predicates drawn from a limited dictionary called primitives. With a limited dictionary, more predicates are needed to distinguish most concepts, rather than a symbol per concept representation, but more details may be captured. Processing may efficiently manipulate analogies and common assertions from generic expressions that apply to more specific ones. Graphs may depict the logical expressions for human manipulation. A web site (http://bobkirby.no-ip.info/) allows collaboration in establishing primitives and logical expression graphs, particularly for word meanings

Speaker Bio

Bob Kirby received a Ph.D. degree in Computer Science from the University of Maryland, College Park. He continues an independent project in Knowledge Representation using his commercial background in Artificial Intelligence (Image Understanding), Expert Systems, and software development. He pursues a machine representation of common sense knowledge and natural language semantics, which is different from those typically used for the semantic web. Yet he also looks to help, as an employee (http://www.linkedin.com/in/bobkirby), with existing research and development in natural language processing such as search or in general software development with Lisp (Clojure), Java, or C++.

Event page provided by ACM