Interactive Probabilistic Search for GikiCLEF

In this paper we will briefly describe the approaches taken by the Berkeley Cheshire Group for the GikiCLEF task of the QA track. Because the task was intended to model some aspects of user search, and because of the complexity of the topics and their geographic elements, we decided to conduct interactive searching of the topics and selection of results. Because of the vagueness of the task specification early-on, some disagreements about what constituted a correct answer, and time constraints we were able to complete only 22 of the 50 topics. However, in spite of this limited submission the interactive approach was very effective and resulted in our submission being ranked third overall in the results. Categories and Subject Descriptors H.3 [Information Storage and Retrieval]: H.3.1 Content Analysis and Indexing; H.3.3 Information Search and Retrieval General Terms Algorithms, Performance, Measurement Keywords Cheshire II, Logistic Regression

• Type: Conference
• Year: 2009
• Where: CLEF
• Authors: Ray R. Larson
• Keywords: Berkeley Cheshire Group, CLEF 2009, GikiCLEF Task, Information Technology, Retrieval General Terms