We present an approach to query mod­el­ing that uses the tem­po­ral dis­tri­b­u­tion of doc­u­ments in an ini­tially retrieved set of doc­u­ments. Such dis­tri­b­u­tions tend to exhibit bursts, espe­cially in news related doc­u­ment col­lec­tions. We hypoth­e­size that doc­u­ments in those bursts are more likely to be rel­e­vant than oth­ers. Pred­i­cated on this, we expand queries with the most dis­tin­guish­ing terms in high qual­ity doc­u­ments sam­pled from bursts. We show how the most com­monly used decay func­tion for recent doc­u­ment retrieval can be used as prob­a­bilis­tic model for tem­po­ral retrieval in gen­eral. The effec­tive­ness of our mod­els is demon­strated on both news col­lec­tions and a col­lec­tion of blog posts.

  • [PDF] M. Peetz, E. Meij, M. de Rijke, and W. Weerkamp, “Adap­tive Tem­po­ral Query Mod­el­ing,” in ECIR ’12, 2012.
    [Bib­tex]
    @inproceedings{ECIR:2012:peetz,
  Author = {Peetz, Maria-Hendrike and Meij, Edgar and de Rijke, Maarten and Weerkamp, Wouter},
  Booktitle = {ECIR '12},
  Date-Added = {2011-11-23 18:10:40 +0100},
  Date-Modified = {2011-11-23 18:11:44 +0100},
  Title = {Adaptive Temporal Query Modeling},
  Year = {2012}}