... pom l =p(wom ,l |zom= l, com, v l ) represents the probabil-ity over all words v l assigned to topic zomat level l, by sampling only for words in wom ,l . Similarly,psn ,l = ... sen-tence level building on hierarchical Latent Dirich-let Allocation (hLDA) (Blei et al., 2003a), whichis a generalization of LDA (Blei et al., 2003b). Weconstruct a hybrid learning algorithm ... ma-chines. In NIPS 9, 1997.A. Haghighi and L. Vanderwende. Exploring con-tent models for multi-document summarization. In NAACL HLT-09, 2009.T. Joachims. Making large-scale svm learningpractical....