%0 Conference Paper %1 citeulike:580822 %A Kubica, Jeremy %A Moore, Andrew %A Schneider, Jeff %A Yang, Yiming %B AAAI '02: Eighteenth national conference on Artificial intelligence %C Menlo Park, CA, USA %D 2002 %I American Association for Artificial Intelligence %K community %P 798--804 %T Stochastic link and group detection %U http://portal.acm.org/citation.cfm?id=777092.777215 %X Link detection and analysis has long been important in the social sciences and in the government intelligence community. A significant effort is focused on the structural and functional analysis of "known" networks. Similarly, the detection of individual links is important but is usually done with techniques that result in “known” links. More recently the internet and other sources have led to a flood of circumstantial data that provide probabilistic evidence of links. Co-occurrence in news articles and simultaneous travel to the same location are two examples. We propose a probabilistic model of link generation based on membership in groups. The model considers both observed link evidence and demographic information about the entities. The parameters of the model are learned via a maximum likelihood search. In this paper we describe the model and then show several heuristics that make the search tractable. We test our model and optimization methods on synthetic data sets with a known ground truth and a database of news articles. %@ 0262511290

@inproceedings{citeulike:580822, abstract = {Link detection and analysis has long been important in the social sciences and in the government intelligence community. A significant effort is focused on the structural and functional analysis of "known" networks. Similarly, the detection of individual links is important but is usually done with techniques that result in “known” links. More recently the internet and other sources have led to a flood of circumstantial data that provide probabilistic evidence of links. Co-occurrence in news articles and simultaneous travel to the same location are two examples. We propose a probabilistic model of link generation based on membership in groups. The model considers both observed link evidence and demographic information about the entities. The parameters of the model are learned via a maximum likelihood search. In this paper we describe the model and then show several heuristics that make the search tractable. We test our model and optimization methods on synthetic data sets with a known ground truth and a database of news articles.}, added-at = {2006-09-25T12:54:00.000+0200}, address = {Menlo Park, CA, USA}, author = {Kubica, Jeremy and Moore, Andrew and Schneider, Jeff and Yang, Yiming}, biburl = {https://www.bibsonomy.org/bibtex/23f3e2e39b469e5d15bef2df25ab186de/grahl}, booktitle = {AAAI '02: Eighteenth national conference on Artificial intelligence}, citeulike-article-id = {580822}, comment = {probabilistic model (based on naive bayes) support non-exclusive group membership Questions addressed: + List all members of G1. + List all the groups for which E1 and E2 are both members. + List a set of suspected aliases (entities that are in th same group(s) but never appear in the same link). They do not follow the mixture model principle, because with unlimited much evidence, only exclusive membership is found. --> Note, that the aspect model by Hofmann/Cohn does not have this flaw. They model noisy links exclicitly in two flavors: + Innocent Link Assumption (completely random links) + Innocent Link Member Assumption (group generated links that contain random non-group members) Their probabilistic model includes nodes for + Demographic Data (observed attributed of each actor) + Demographic Model (p(Member Gi|DD)) + Chart (1-0 membership of actors in groups) + Link Model (PInnocence1, PInnocence2 for each link type) + Link Data (lists observed coocurrence of actors with the link type) Experiments They applied their model to research interests listed on researchers webpages. In the first experiment, they considered researchers as actors and common interests as links, hoping to find project teams, which was not sooo successful. In the second experiment they consideres interests as actors and two interests of the same researcher as linked, which led to good descriptions of the interest field.}, interhash = {7a471a32a59e73c43dc0dd64d55176d2}, intrahash = {3f3e2e39b469e5d15bef2df25ab186de}, isbn = {0262511290}, keywords = {community}, month = {August}, pages = {798--804}, priority = {0}, publisher = {American Association for Artificial Intelligence}, timestamp = {2006-09-25T12:54:00.000+0200}, title = {Stochastic link and group detection}, url = {http://portal.acm.org/citation.cfm?id=777092.777215}, year = 2002 }