Using Stream Features for Instant Document Filtering
Abstract
In this paper, we discuss how event processing technologies can be employed for real-time text stream processing and information filtering in the context of the TREC 2012 microblog task. After introducing basic characteristics of stream and event processing, the technical architecture of our text stream analysis engine is presented. Employing well-known term weighting schemes from document-centric text retrieval for temporally dynamic text streams is discussed next, giving details of the ESPER Event Processing Agents (EPAs) we have implemented for this task. Finally, we describe our experimental setup, give details on the TREC microblog runs as well as the result thereafter with our system including some extensions and give a short interpretation of the evaluation results.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2012
- Accession Number
- ADA580692
Entities
People
- Andreas Bauer
- Christian Wolff
Organizations
- University of Regensburg