Accession Number : ADA618576


Title :   Drexel at TREC 2014 Federated Web Search Track


Descriptive Note : Conference paper


Corporate Author : DREXEL UNIV PHILADELPHIA PA COLL OF COMPUTING AND INFORMATICS


Personal Author(s) : Zhao, Haozhen ; Hu, Xiaohua


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/a618576.pdf


Report Date : Nov 2014


Pagination or Media Count : 6


Abstract : This paper reports our participation in the Federated Web Search Track in TREC 2014. We submitted 21 runs for all the three tasks: Vertical Selection (7), Resource Selection (7) and Results Merging (7). Our main purpose is to test several established resource selection methods on the new realistic FedWeb test collections. We evaluated 7 well known resource selection methods for the vertical selection and resource selection tasks. The effectiveness of these methods in the RS tasks does not carry to the VS tasks, which implies that more sophisticated algorithms and more diverse sources of evidence are needed for solving the VS task effectively. Our Results Merging experiments reveal the correlation between the performance of RM and the performance of its input RS results.


Descriptors :   *DATA MINING , *INFORMATION RETRIEVAL , CLASSIFICATION , COMPUTERIZED SIMULATION , CORRELATION , EXPERIMENTAL DESIGN , INTERNET , KNOWLEDGE MANAGEMENT , MARKOV PROCESSES , MATHEMATICAL MODELS , PATTERN RECOGNITION , RANK ORDER STATISTICS , SEMANTICS


Subject Categories : Information Science


Distribution Statement : APPROVED FOR PUBLIC RELEASE