Information retrieval using very short Krylov sequences
Katarina Blom, Axel Ruhe · 2001
Introduction The task is to search among the documents in a large data base for those that contain information of interest stated in a query. We summarize the contents of the data base as an m \\Theta n term document matrix X , where each column represents one document and each row one term, in the simplest case just a specific word. An element x ik is nonzero whenever term i is present in document k. The query is now q, an m vector of terms, and we may form the scalar product p T = q T X (1) to get p a choice vector whose nonzero elements indicate which of the documents that contain any of the terms in the query. This is the way one looks at it in the vector space methods of inform