Pitt at TREC 2006: Identifying Experts via Email Discussions
暂无分享,去创建一个
Identifying experts in a certain domain or a subject area has always been a challenge in various settings including commercial, academia, and governmental institutions. Our interests in this year's TREC Enterprise track are to utilize the email communications as the basis for identifying experts and their expertise on certain topics. In this report, we presented a method for identifying experts based on the emails they sent around. We hypothesize that experts would be more active in relevant email threads, would send longer emails, and would participate in the discussion at the very beginning of the threads. An algorithm based on these hypotheses was developed and tested in this year TREC enterprise track experiments to find experts for 49 topics based on documents in the W3C collections. Our initial experiment results produced suboptimal performance. This motivated us to examine the hypotheses more closely in the context of provided ground truth. Interestingly, the analysis on ground truth seems to confirm that all of our hypotheses have their merits in finding experts, so one future important question is how to utilize these rules in a right way.
[1] Paul P. Maglio,et al. Expertise identification using email communications , 2003, CIKM '03.
[2] M. de Rijke,et al. Formal models for expert finding in enterprise corpora , 2006, SIGIR.
[3] Nick Craswell,et al. Overview of the TREC 2005 Enterprise Track , 2005, TREC.
[4] Nick Craswell,et al. Overview of the TREC 2006 Enterprise Track , 2006, TREC.