Spectral-based document retrieval

The fast vector space and probabilistic methods use the term counts and the slower proximity methods use term positions. We present the spectral-based information retrieval method which is able to use both term count and position information to obtain high precision document rankings. We are able to perform this, in a time comparable to the vector space method, by examining the query term spectra rather than query term positions. This method is a generalisation of the vector space method (VSM). Therefore, our spectral method can use the weighting schemes and enhancements used in the VSM.