Research in Knowledge Representation, Reasoning, and Acting Stuart C. Shapiro Professor, CSE Affiliated Professor, Linguistics, Philosophy Director, Center for Cognitive Science Director, SNePS Research Group ACM Distinguished Scientist Fellow, AAAI Faculty Member: Interdisciplinary MS in Computational Linguistics Center for MultiSource Information Fusion National Center for Ontological Research National Center for Geographic Information and Analysis Long-Term Goal • Theory and Implementation of Natural-Language-Competent Computerized Cognitive Agent/Robot • and Supporting Research in Artificial Intelligence Cognitive Science Computational Linguistics. Research Overview '06 S. C. Shapiro 2 Cassie • A computational cognitive agent – Embodied in hardware – or software-simulated – Based on SNePS and (M)GLAIR. Research Overview '06 S. C. Shapiro 3 MGLAIR Agent Architecture Mind Body KL (SNePS) Independent of lower-body implementation PMLa PMLb I/P s o c k e t s Dependent on lower-body implementation PMLc Proprioception Speech W O R L D Hearing SAL Vision Motion Research Overview '06 S. C. Shapiro 4 SNePS SNePS is a Logic-Based Frame-Based Network-Based knowledge representation, reasoning, and acting system. Research Overview '06 S. C. Shapiro 5 Some Important SNePS Features • First-person beliefs – Not third-person “truth” about agent or world • Beliefs are current beliefs – Even if about the past • On-line acting • Reified propositions as well as acts & … • Neither states nor times are privileged Research Overview '06 S. C. Shapiro 6 Some Recent & Current Research Projects • A General Characterization of Answers to Questions • KRR for Information Fusion for Cyber Security • Information Extraction for Soft Target Exploitation and Fusion • Ontological reasoning for GeoCLEF: multilingual geographic information retrieval • Intermedia Performance Studio – Actor-agents in VR drama Research Overview '06 S. C. Shapiro 7 A General Characterization of Answers to Questions • Every clause descended from a query clause in resolution theorem proving is an answer to the query. • General form of an answer: [H ] [G ] Q H is Hypothetical component (optional) G is Generic component (optional) Q is Question component (either generic or specific) • Example: {~Cat(Boots), ~Tuna(x), Answer(Likes(Boots, x))} Cat(Boots) x (Tuna(x) Likes (Boots, x)) If Boots is a cat, then Boots likes to eat any tuna. [D. T. Burhans & S. C. Shapiro, Defining Answer Classes Using Resolution Refutation, Journal of Applied Logic 5,1 (March 2007), in press.] Research Overview '06 S. C. Shapiro 8 KRR for Information Fusion for Cyber Security • Use SNePS to reason about computer networks, and about potential, and actual attacks. • Provide ontology-oriented SNePS GUI • A task of the National Center for Multisource Information Fusion • With Moises Sudit (IE & CMIF) & Michael Kandefer (CSE: RA) • Funded by the Air Force Research Laboratory (AFRL) Research Overview '06 S. C. Shapiro 9 SNePS Ontology GUI • SNePS GUI supports the loading and exporting of SNePS files in several formats. • Allows the display of binary predicates as a tree hierarchy Research Overview '06 S. C. Shapiro 10 Selecting a Predicate for the Tree View • The user selects a predicate from the drop down menu, after which the tree is generated Research Overview '06 S. C. Shapiro 11 Class Hierarchy • This view shows a small class hierarchy for hosts on a network Research Overview '06 S. C. Shapiro 12 Part Of • This view shows a part of hierarchy. Research Overview '06 S. C. Shapiro 13 Network View • • • • Shows the network representation (drawn using the JUNG network visualization tool) of a SNePS belief base User can zoom, move nodes, and save images as JIMI supported file types (jpg, bmp, png) Tooltips provide information about the propositions expressed by the nodes in the network Ex. This view shows assertions about the connectivity of various hosts (h0-h4) Research Overview '06 S. C. Shapiro 14 Information Extraction for Soft Target Exploitation and Fusion • Review relevant NLP tools. • Translate unstructured text messages to structured summaries. • E.g. From 03/17/07 Sabah Dulaimi arrived at 1035. To (m2 (date (m1 (day 17) (month 03) (year 07)))) (m3! (act arrive) (agent Sabah Dulaimi) (time 1035)) • • With Shane Axtell (CL: RA) Funded by the US Army CERDEC (Communications-Electronics Research, Development and Engineering Center) Research Overview '06 S. C. Shapiro 15 Ontological Reasoning for GeoCLEF • Provide additional search terms to retrieve articles for queries such as International sports competitions in the Ruhr area and provide means of eliminating irrelevant answers. • With Miguel Ruiz (LIS), June Abbas (LIS), Thomas Bittner (PHI & GEO), & David Mark (GEO). Research Overview '06 S. C. Shapiro 16 Intermedia Performance Studio • Buffalo-Region resource focusing on the integration of live actors, virtual avatars, intelligent actor-agents, dynamic sets and live, mobile audience members. • With Sarah Bay-Cheng (THD & DMS), Josephine Anstey (DMS), David Pape (DMS), & Jon Bona (CSE: RA) • Funded by UB Provost’s Interdisciplinary Research Development Fund (IRDF) • See http://vrstudio.buffalo.edu/ips/wiki/ Research Overview '06 S. C. Shapiro 17 The Trial The Trail • Virtual Reality Drama with SNePS/GLAIR agent-actors. • With Josephine Anstey (DMS), David Pape (DMS), and CSE grads: Jon Bona, Albert Goldfain, Mike Kandefer, Vishwac Sena Kannan, Madhumitha Nagarajan Research Overview '06 S. C. Shapiro 18 The Trial The Trail Bad guy agents hassling a human participant Research Overview '06 S. C. Shapiro 19 For More Information • Shapiro: http://www.cse.buffalo.edu/~shapiro/ • SNePS Research Group: http://www.cse.buffalo.edu/sneps/ – Meets Tuesdays 10-12, 242 Bell Hall – Join us! • Register for CSE563 in Spring, 2007 Research Overview '06 S. C. Shapiro 20