Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!mcsun!ukc!dcl-cs!aber-cs!odin!pcg From: pcg@odin.cs.aber.ac.uk (Piercarlo Grandi) Newsgroups: comp.databases Subject: Re: >Fault-tolerant Information Recall Message-ID: Date: 14 Apr 90 13:31:59 GMT References: <1990Apr3.200220.9513@sctc.com> <22782@netnews.upenn.edu> <1990Apr4.204241.2249@sctc.com> Sender: pcg@aber-cs.UUCP Organization: Coleg Prifysgol Cymru Lines: 61 In-reply-to: endrizzi@sctc.com's message of 4 Apr 90 20:42:41 GMT In article <1990Apr4.204241.2249@sctc.com> endrizzi@sctc.com (Michael Endrizzi ) writes: This depends on the application. If I am using Unix and I type in the command "zgrep" instead of "grep", it would be great if the command parser could say "I ain't got no 'zgrep', but here are a list of commands that are approximately close to 'zgrep'". Already Exists! Try the 'tcsh' shell with the '$' command (well, it actually corrects the spelling for you; you can then change it if you don't like it). Some applications would consider this command parser fault-tolerant because it is able to determine that a fault occurs and not return the empty set, but instead attempt to recover. Other applcations would consider this a nuisance and not fault-tolerant. (Works great on Emacs though). I totally disagree with you calling this fault-tolerant: this has a very specific meaning in CS circles, that processing continues in the presence of faults. What you are discussing is known as 'fuzzy retrieval'. Actually, looking at the other articles in this thread, there are two distinct, but interoperable, technologies that people; one if fuzzy retrieval, the other is "deductive retrieval". The first is often found in bibliographic databases, where for example you want to ask fuzzy queries about a subject, the second is in a sense the definition of expert system. Fuzzy retrieval can be based on a many-valued logic called fuzzy logic, in which predicates are not true or false, but have a weight between 0 and 1. In bibliographic databases the predicate you want to use is 'relevant to my query', and then each document is indexed not just by a set of keywords, but also by a weight/relevance factor per each subject you may want to index on. Note that fuzzy databases are not the same as statistical databases, even if there are obvious similarities. Deductive retrieval is the idea that the sum of facts stored in a database contains also the facts that can be derived, according to some inference rules, from those explicitly stored in it. Many so called expert systems are just deductive databases. You can also have fuzzy deductive retrieval, in which the inference rules use fuzzy instead of classical logic, and weights as truth values. Some issue of 'The Economist' ventured to predict that one of the biggest software businesses to come will be providing intelligent, i.e. fuzzy, statistical and deductive, database navigators, as even now the problem of extracting interesting data from huge, amorphous corporate, commercial or governmental databases is pretty tough, and as data masses, things can only get worse. There is periodic conference on so called 'intelligent databases', and I think reading the proceedings is a good sport. -- Piercarlo "Peter" Grandi | ARPA: pcg%cs.aber.ac.uk@nsfnet-relay.ac.uk Dept of CS, UCW Aberystwyth | UUCP: ...!mcvax!ukc!aber-cs!pcg Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk