Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!ukc!edcastle!edcogsci!rda From: rda@cogsci.ed.ac.uk (Robert Dale) Newsgroups: comp.sys.mac.apps Subject: Re: Grammar Checker Message-ID: <4140@scott.ed.ac.uk> Date: 20 Feb 91 15:56:22 GMT References: Organization: Centre for Cognitive Science, Edinburgh, UK Lines: 53 Oh dear I knew this would happen ... rq02+@andrew.cmu.edu (Richard Quadrel) writes in response to my statement that >> The only other Mac grammar checker is Grammatik, but Correct Grammar >> is far superior. >I don't know what kind of papers Mr. Dale writes, so I have no idea why he >made that comment about Sensible Grammar. While no grammar checker is >perfect, both Sensible Grammar and Grammatik were successful in dicovering >errors in both my technical papers and informal correspondance. ... >... >Sensible Grammar doesn't seem to catch as many potential problems as >Grammatik does, but it's still a pretty good program. And either of these >beats the pants off of Correct Grammar, in my opinion. For Mr. Dale and >others: try running Correct Grammar and Grammatik on a sample file (I'll >be happy to provide one of my own!) and compare for yourself! Ok. I was addressing only the issue of *syntactic error detection and correction* -- the original poster asked for a *grammar* checker. These programs do other things, but Sensible Grammar isn't particularly interested in doing *syntax* (= grammar in my language) checking -- I'm talking here about things like subject-verb number agreement and other genuine syntactic problems. Because of its limited pattern matching techniques, Sensible Grammar generates a lot of false positives: for example, if you say "When is she going to work?", SG will suggest replacing "she going" by "her going". Grammatik (at least in Version 3; haven't tested Version 4 yet) uses (as far as we can tell by trying to break it) technology based on the old Unix WWB PARTS program, plus some error-locating heuristics on top. This is pretty limited for catching syntactic errors: for example, it doesn't spot the problem with "John and Mary starts the car". Correct Grammar, on the other hand, does appear to carry out some real parsing of the text. These programs all do many other things quite well, and I would be the first to agree that the interface is very important, although there are of course different preferences here. But when it comes to detecting genuine syntactic problems, Correct Grammar wins. I should add that (a) my awareness of the limitations of various systems here is largely due to the detailed examinations performed by a colleague, Shona Douglas, and that (b) I have no connections with the manufacturers of any of these programs. R -- Robert Dale Phone: +44 31 650 4416 | University of Edinburgh UUCP: ...!uunet!mcvax!ukc!its63b!cogsci!rda | Centre for Cognitive Science ARPA: rda%cogsci.ed.ac.uk@nsfnet-relay.ac.uk | 2 Buccleuch Place JANET: rda@uk.ac.ed.cogsci or R.Dale@uk.ac.ed | Edinburgh EH8 9LW Scotland