abstract |
|
In Information Retrieval(IR) a central problem is determining the
importance of a given datum to a query. Moreover, the problem of
determining importance of elements in a collection is central to a
variety of areas related to IR like applied graph theory, clustering,
text analysis or network analysis. Spectral techniques using matrix
representations of data offer an intuitive approach to measure
importance and have been successfully applied in IR. After a review of
fundamental algebraic properties, the talk sketches some influential
applications for link, network and text analysis to exemplify the power
of spectral analysis. In addition, a combined spectral method for
analysis and visualization of text documents is outlined.
|