Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Description and performance analysis of signature file methods

Christodoulakis Stavros, Christos Faloutsos

Full record


URI: http://purl.tuc.gr/dl/dias/BA5E9D2B-D9DD-46CC-ABA2-A38BC3AFF5B9
Year 1987
Type of Item Peer-Reviewed Journal Publication
License
Details
Bibliographic Citation S. Christodoulakis., C. Faloutsos ,"Description and performance analysis of signature file methods ",ACM Trans. on Inf. Syst.,vol.3,no.3,pp.237-257 ,1987.doi :10.1145/27641.28057 https://doi.org/10.1145/27641.28057
Appears in Collections

Summary

Signature files have attracted a lot of interest as an access method for text and specifically for messages in the office environment. Messages are stored sequentially in the message file, whereas their hash-coded abstractions (signatures) are stored sequentially in the signature file. To answer a query, the signature file is examined first, and many nonqualifying messages are immediately rejected. In this paper we examine the problem of designing signature extraction methods and studying their performance. We describe two old methods, generalize another one, and propose a new method and its variation. We provide exact and approximate formulas for the dependency between the false drop probability and the signature size for all the methods, and we show that the proposed method (VBC) achieves approximately ten times smaller false drop probability than the old methods, whereas it is well suited for collections of documents with variable document sizes.

Services

Statistics