- From: Martin J. Dürst <mduerst@ifi.unizh.ch>
- Date: Fri, 10 Oct 1997 11:19:07 +0100 (MET)
- To: Graham Klyne <GK@acm.org>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
On Thu, 9 Oct 1997, Graham Klyne wrote: > In private discussions, it has been suggested to me that language issues > are a poor basis upon which to develop a "quality" rating system -- there > is just too much subjectivity involved. Thanks for bringing up these private discussions. There could be various issues of subjectivity you are alluding to. One thing is that some people are very convinced that languagage, in various ways, is very important, more important than other things, and that this can affect some discussions. The second thing is that language abilities, and therefore preferences, differ widely for each individual, so this would mean that language is a particularly good example for the preference side, if we need something where preferences are clearly varying. Of course, for your problem of granularity, we would need something where preferences have only fine distinctions. The third thing I could think of that you are alluding to is that the rating of the quality of a document with respect to language is very subjective, i.e. there are widely differing oppinions about what is "good English", "good French", and so on. In this respect, I think I would have to oppose; other quality things such as typographic design (in formats where this matters), document structure, audio quality, and so on, may also easily be subject to such subjectivities. There may be other aspects of subjectivity in language, I would like to hear about them. In conclusion, I guess it is fair to say that a "quality" rating system that is developed ONLY on the basis of language issues will not be a good solution. But a "quality" rating system that ignores language issues will probably be as bad, if not worse. > >Various aspects of quality are usually combined by multiplication; > [...] > > It was a *premise* of my posting (one with which you may well disagree) > that the quality factors were used to simply rank alternatives. It had > been suggested to me in offline discussion that other document selection > systems which attempted to perform arithmetic manipulation of quality > factors gained limited benefit from such manipulations (unfortunately, I > don't remember details of the example system offered). I can very well agree with your statement. The examples we usually see, including the ones I made, are idealistic toy examples. The underlying problem is that even for the human user, it's difficult to decide whether French Postscript or English HTML is preferred. As long as the quality difference isn't above a certain threshhold, there is probably not much of a real preference. Also, there might be some drawing in Postscript that doesn't show well in HTML, or some important link or form in HTML that isn't as convenient in Postscript, and the human end user may not know what is more important to him/her until he has looked at one (or both!) versions of the document. So how should the arithmetic be able to figure this out :-? Regards, Martin.
Received on Friday, 10 October 1997 02:38:22 UTC