[Fwd: Suggestion for Improving Search at PHP.NET]

Hi folks, This is a message I received yesterday regarding possible work to the PHP site's search functionality. The author has a history of search tech work and makes some good points, and I tend to agree with him that the PHP site's searching could use some work. Anyway, I don't have time at the moment to deal with this and I don't work on the site anyway, so I'm throwing it out there for discussion. Thoughts? Torben -----Forwarded Message----- > From: J. Scott Johnson <scott@fuzzygroup.com> > To: torben@php.net > Subject: Suggestion for Improving Search at PHP.NET > Date: 17 Mar 2002 11:50:56 -0500 > > Hi Torben, > > I wasn't sure who to send this to but since I have talked to you at least > once and I know you have some relationship to the documentation, I thought > I'd send it to you. Sorry if its a bother or off point for you. > > I'm (fairly rapidly) climbing the learning curve for PHP and what I am > finding is that my search queries often don't easily match up to functions > in the manual. For example I searched for "writing to start of file" (no > quotes) and got a fairly arcane set of responses including one on writing > PDFs. While this is algorithmically correct from the indexer's perspective, > from a person's perspective it actually isn't. I'd wager that people use > the PDF writing abilities rarely as compared to simply writing text files > and the like. > > My experience has been that if you analyze your http logs, you could extract > the search entries and probably find that a large number of searches are > actually a relatively small number of common queries. For any language, its > always the same issues -- writing files, strings, etc. One way to improve > this is to use a query preprocessor to add a level of human intelligence to > your search. This works well when you have a fairly constant url structure. > What you do is manually map queries to pages and then change your result > list to be a two tier hierarchy (Human Answers and Documents). Thus, your > results page becomes something like this: > > For your search, {insert query here}, we found: > > Human Recommended Answers > a.. search result > b.. search result > Documents We Found: > a.. search result > b.. search result > To make this a lot easier, get the users of the site to contribute back > their problems and the webpage which answered it. I suspect the user > community would be happy to help. > > Note I have a pretty serious background (10 + years) in search and retrieval > http://www.fuzzygroup.com/about/sjohnsonresume.htm . I do know > what I am > talking about here. This is actually pretty much the technique used by > AskJeeves (which I don't like as a general tool but does work quite well in > focused knowledge domains such as a language). > > If I can help at all with this, happy to... > > Scott > J. Scott Johnson > > Physical > * * * * * * * * * * * * * * * * * * * * * * * * * * > 80 Spring Road > Nahant, MA 01908 > > Phone: > * * * * * * * * * * * * * * * * * * * * * * * * * * > 781 592 0262 - home > 617 970 4719 - cell > > Virtual: > * * * * * * * * * * * * * * * * * * * * * * * * * * > sjohnson@fuzzygroup.com > http://www.fuzzygroup.com/ > Yahoo IM: fuzzygroup > >