KOJAK Group Finder: Scalable Group Detection via Integrated Knowledge-Based and Statistical Reasoning
Abstract
This report describes the KOJAK Group Finder which is a highly scalable, hybrid logic-based statistical link discovery system designed to solve group detection problems. The system takes primary and secondary evidence as input and produces group hypotheses with ranked lists of group members as output. Under funding of this project, the Group Finder was developed from a research prototype into a mature software component that can be transitioned into operational environments. Among the many improvements, scalability and processing speed was improved by over a factor of 10, the software is now available in C++ and Java versions for Linux, as well as Windows platforms and now also supports Oracle databases in addition to MySQL. A Java-based GUI was added to facilitate usage by non-experts.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2006
- Accession Number
- ADA457191
Entities
People
- Hans Chalupsky
- Jafar Adibi
- Thomas A. Russ
Organizations
- University of Southern California