Document Retrieval Systems,

Some of the mathematical properties of a term matching information retrieval system are investigated. It is shown that the common retrieval method of using a query vector, a matching function, and a threshold is equivalent to retrieving documents by requiring that a specific mathematical combination of the over and under indexing errors between the query vector and document index vector is bounded. Furthermore, the over and under indexing error set provides a sample space for a probabilistic description of the retrieval process. Using this approach an explicit form of the expected recall ratio is derived. Author