|
Finding Identity Group "Fingerprints" in Documents
August 2008
Lashon B. Booker, The MITRE Corporation
ABSTRACT
This paper describes how social identity group "fingerprints" can be extracted from a document collection by applying topic analysis methods in a novel way. The results of document classification experiments suggest that these group-level attributes provide better predictions of group affiliation than document-level attributes. Applications of this method for forensic authorship analysis are also discussed.
Additional Search Keywords
topic analysis, latent dirichlet allocation, document classification, authorship analysis
|
