%0 Conference Paper %1 citeulike:396067 %A Cohn, David %A Chang, Huan %B ICML '00: Proceedings of the Seventeenth International Conference on Machine Learning %C San Francisco, CA, USA %D 2000 %I Morgan Kaufmann Publishers Inc. %K taxonomy socialnets %P 167--174 %T Learning to Probabilistically Identify Authoritative Documents %U http://portal.acm.org/citation.cfm?coll=GUIDE&dl=GUIDE&id=657957#

@inproceedings{citeulike:396067, added-at = {2006-06-16T10:34:37.000+0200}, address = {San Francisco, CA, USA}, author = {Cohn, David and Chang, Huan}, biburl = {https://www.bibsonomy.org/bibtex/2d519bd3f3c447e24bb55f414740fb19e/ldietz}, booktitle = {ICML '00: Proceedings of the Seventeenth International Conference on Machine Learning}, citeulike-article-id = {396067}, comment = {Method: tempered EM (like Hofmann) "In this paper we introduce a bibliometric method for identifying hubs and authority documents based on this probabilistic model of citation." Mathematical analogue to Hofmanns pLSA paper. Document-citation pairs (d,c) (C == D, but different meaning) <li> P(d,c)=P(d)*P(c|d) and <li> P(c|d)= SUM_z (P(c|z) * P(z|d)) total likelihood of observed citation matrix M = L(M) = PROD_(d,c) (P(d,c)) Tasks: <LI> authority of a document P(c|z) <LI> in which community is a cited document? P(z|c) (calculate via bayes rule) <LI> which documents are characteristic? char_ij = P(z_j|c_i)*(Pc_i|z_j) <LI> Which documents are hubs? P(d|z) <LI> Document membership P(z|d) <LI> Document characteristic probability P(z|d)*P(d|z) Problem with pHITS: although it gives promising results on citation graphs, the famous jaguar football/car example does not get reflected. Future work: <li> combine pHITS with pLSA (done in http://www.citeulike.org/user/ldietz/article/391312) <li> Use PCA calculations as seeding and estimate number of factors (topics) <li> calculate hierarchy with split strategy}, interhash = {78901a7619e22c948d3d26e39ecfaa5b}, intrahash = {d519bd3f3c447e24bb55f414740fb19e}, keywords = {taxonomy socialnets}, pages = {167--174}, priority = {0}, publisher = {Morgan Kaufmann Publishers Inc.}, timestamp = {2006-06-16T10:34:37.000+0200}, title = {Learning to Probabilistically Identify Authoritative Documents}, url = {http://portal.acm.org/citation.cfm?coll=GUIDE\&#38;dl=GUIDE\&#38;id=657957#}, year = 2000 }