Description and performance analysis of signature file methods for office filing

Reviewer: Dennis Roy Thompson

Christos and Stavros stuck in their thumbs and pulled out a signature file method plum. Access method classifications are full text scanning, inversion, signature files, clustering, and multi-attribute hashing. Signature file methods are concerned with creating a terse encoded index from an abstract of the text. Both the text and the “signature” are then just appended to their respective files. To search, the entire index must be scanned and comparisons and/or calculations made according to the search criteria. The object is to create a signature that is terse, never misses, and has few false matches (false drops). The methods that the authors look at are Word Signature (WS), Superimposed Coding (SC), Bit-block Compression (BC), Run Length encoding (RL), and Variable Bit-block Compression (VBC). They present a detailed design of the BC method and a proposal of the VBC method with supporting derivations. I would like to see them do research in combinations of access methods such as using daily (or some reasonable length of time) signature methods with batching into inversion. Also, the authors should not assume that the average user is not familiar with Boolean logic or the authors may become familiar with Boolean egg. I've seen too many users use it without knowing its name.