The search for meaning in large text databases
Daniel Gruhl, Marvin Minsky · DSpace@MIT (Massachusetts Institute of Technology) · 2000
The task of retrieving information is becoming increasingly complicated as larger coropa are searched. Systems that employ single-pass statistical methods often swamp the user with an excessive number of hits or fail to find any exact matches and return none. This thesis examines a system that employs richer representation of the texts in its corpus. This understanding is developed by a collection of experts that examine articles in the database, looking at both the text itself and the annotations placed by other agents. These agents employ a variety of methods ranging from statistical examination to natural-language parsing to query expansion through specific-purpose knowledge bases. Retrieval servers use these augmented feature sets to provide more useful responses to user queries. Thesis Chairman: Marvin Minsky Title: Toshiba Professor Of Media Arts And Sciences 2 Acknowledgments There are many people who deserve thanks for helping to make this thesis a reality. First and forem...