Using Nonnegative Matrix and Tensor Factorizations
                        for Topic Detection and Tracking

                                Michael W. Berry

           Department of Electrical Engineering and Computer Science
                       University of Tennessee, Knoxville 

Automated approaches for the identification and clustering of semantic features
or topics are highly desired for text mining applications.  Using low rank
non-negative matrix factorizations (NNMFs) to retain natural data
non-negativity, one can eliminate subtractive basis vector and encoding
calculations present in techniques such as principal component analysis for
semantic feature abstraction.  Moving beyond two-way factorizations, we
demonstrate how non-negative tensor factorizations (NNTFs) can be used to
capture temporal and semantic proximity and thereby enable the tracking of both
targeted and latent (previously unknown) discussions or communication patterns.
Demonstrations of NNMF and NNTF algorithms for topic (or discussion) detection
and tracking using the Enron Email Collection and documents from the Airline
Safety Reporting System (ASRS) are provided.  Recent work with Prof. Amy N.
Langville (College of Charleston) on alternative $L_p$ norm formulations of the
objective functions associated with NNMF factorization will also be presented.

****************************************************************************

Short Bio:

Michael W. Berry holds the title of Full Professor and Associate Department
Head in the newly formed Department of Electrical Engineering and Computer
Science at the University of Tennessee, Knoxville.  He first joined the
Department of Computer Science in 1991 as an Assistant Professor and was
promoted to Associate Professor in 1997.  He was later promoted to Full
Professor in 2003.  He served as Interim Department Head of Computer Science
from January 2004 to June 2007.  Prior to UT, he spent 1 year as an Assistant
Professor in the Department of Computer and Information Sciences at the
University of Alabama at Birmingham.  He received the BS degree in Mathematics
from the University of Georgia in 1981 and the MS degree in Applied Mathematics
from North Carolina State University in 1983.  He worked in the Communications
Product Division of IBM in Raleigh, NC for about 1 year before accepting a
research staff position in the Center for Supercomputing Research and
Development at the University of Illinois at Urbana-Champaign.  In 1990, he
received the PhD degree in Computer Science from the University of Illinois at
Urbana-Champaign.

Prof. Berry is the co-author of "Templates for the Solution of Linear Systems:
Building Blocks for Iterative Methods" (SIAM, 1994) and "Understanding Search
Engines: Mathematical Modeling and Text Retrieval, Second Edition" (SIAM, 2005)
and editor of "Computational Information Retrieval" (SIAM, 2001), "Survey of
Text Mining: Clustering, Classification, and Retrieval" (Springer-Verlag, 2003,
2007), and "Lecture Notes in Data Mining" (World Scientific, 2006).  He has
published over 100 refereed journal and conference publications.

He has organized numerous workshops on Text Mining and was Conference Co-Chair
of the 2003 SIAM Third International Conference on Data Mining (May 1-3) in San
Francisco, CA.  He was also Program Co-Chair of the 2004 Co-Chair of the 2003
SIAM Fourth International Conference on Data Mining (April 22-24) in Orlando,
FL.  He is a member of SIAM, ACM, MAA, and the IEEE Computer Society and is
on the editorial board of "Computing in Science and Engineering", "SIAM
Journal of Scientific Computing", and "Statistical Analysis and Data Mining".

His research interests include information retrieval, data and text mining, 
computational science, bioinformatics, and parallel computing.  Prof. Berry is 
currently supported by grants and contracts from the National Science 
Foundation, National Institutes of Health, and the National Aeronautics and 
Space Administration.