Exploiting Intra-document Structures [Ogilvie Callan 2003] D Intuitively,we want to combine all the parts. but give more weights to some parts Title D Abstract +D2 Think about query-likelihood model... Select D and generate a Body-Part1 +D3 query word using D; Body-Part2 p(QID.R)-TIp(w.ID.R) -TIEs(D,ID.R)pOID.R i=1j=1 D "part selection"prob.Serves as weight for D Can be trained using EM Anchor text can be treated as a "part"of a document CCF-ADL at Zhengzhou University,June 25-27,2010Exploiting Intra-document Structures [Ogilvie & Callan 2003] Title Abstract Body-Part1 Body-Part2 … D D1 D2 D3 Dk Intuitively, we want to combine all the parts, but give more weights to some parts Think about query-likelihood model… 1 1 1 ( | , ) ( | , ) ( | , ) ( | , ) n i i n k j i j i j p Q D R p w D R s D D R p w D R = = = = =   “part selection” prob. Serves as weight for Dj Can be trained using EM Select Dj and generate a query word using Dj Anchor text can be treated as a “part” of a document 8 CCF-ADL at Zhengzhou University, June 25-27, 2010