Capturing and Using Patterns for Evidence Detection

Pattern-based analysis of data plays an increasing role in several important applications. In crime prevention (including securities trading, tax fraud, and homeland security) it is being used both to detect evidence of criminal events and to predict threatening activities before they completely mature. In marketing it is being used to assess trends in the aggregate sentiments of populations as well as the preferences of individuals. In epidemiology it is used to assess health trends in populations and provide early warning for epidemics. In these applications the data is typically incomplete and becomes available incrementally over time, and it can often support alternative interpretations, so assessing the quality of the evolving evidence among a set of competing hypotheses is critical. This symposium will bring together researchers from diverse backgrounds, including machine learning, data management, graph theory, link analysis, information retrieval, privacy, automated reasoning, and knowledge representation, to promote advances in acquiring and using patterns for detecting and managing evidence in data.

Topics of particular interest include:

• Learning patterns from data
• Identifying patterns efficiently within massive, structured, or partially structured data
• Matching algorithms for specific data characteristics (inexact matching, etc.)
• Reasoning with patterns (deduction, abduction, induction, disjunction, negation, etc.)
• Representing patterns and hypotheses
• Hypothesis management (monitoring predictions, interactive refinement, etc.)
• Managing conflicting and uncertain data (probabilities, knowledge gaps, etc.)
• Data and reasoning provenance
• Data access issues (privacy, secrecy, proprietary, etc.)
• Applications (homeland security, fraud detection, epidemiology, marketing, etc.)
• Evaluation of pattern analysis systems
• Pattern and hypothesis sharing among tool sets (interlingua, work flows, etc.)
• Data and algorithm characterization
• Equality detection (alias resolution, etc.)
• Group detection (collaboration networks, etc.)

Submissions: Potential participants may submit a technical paper (up to 8 pages), or a short paper (up to 3 pages) in the form of an extended abstract or a description of a proposed demo or poster. Potential participants who are unable to submit a paper are encouraged to submit a one-page statement of interest. Submissions in PDF using AAAI format should be sent to murray@ai.sri.com.

Organizing Committee: Ken Murray (cochair), SRI International (murray@ai.sri.com); Ian Harrison (cochair), SRI International (harrison@ai.sri.com); Fotis Barlos, BAE Systems; Tina Eliassi-Rad, Lawrence Livermore National Laboratory; Henry Goldberg, National Association of Securities Dealers; Seth Greenblatt, 21st Century Technologies, Inc.; Dunja Mladenic, J. Stefan Institute; Robert Popp, Aptima, Inc.; Ben Rode, Cycorp.

Important Dates:

Additional Information