Beefy Boxes and Bandwidth Generously Provided by pair Networks
Just another Perl shrine
 
PerlMonks  

I have a dream ...

by little (Curate)
on Nov 08, 2000 at 17:02 UTC ( #40515=perlmeditation: print w/replies, xml ) Need Help??

sometimes and well, it's nothing special, but more theoretical for now and I guess it could be of help, however.

As there already is the very usefull Q&A section it is still quite hard to find answers when you don't know the magic question to ask.

Once thinking about this problem of substitutional terms and conditions I thought of something like a "smart parser" for the questions.

So what I want it to do is simply to give answers if I have a question, but my question isn't like the others before. So the question would seem to be a new one, just because I omitted a word for example.
So I woukld get thousands of answers for an unsharp question or would get no answer cause the question was to specific.

But I believe that out there in the deep of several Q&A and FAQ sections are the most of the answers already.

The job of such "swiss army knife" should include the following:
  1. parsing present Q&A to find related topics/problems
  2. subparsing of questions and answers like a "study"
  3. build indices on correlating word matches, e.g. Network->config->RAS->ISP
  4. study a new question, to prevent users from posting already answered questions
  5. compare question with the indexed ones using thesaurus and alternative language expressions, phrases or constructs, e.g. "How to"<->"(How)? can I"
  6. if no matches found ask to redefine the question or to comfirm that it gets posted
  7. if too many matches found, try to break down to most matching while giving the user the possibility to choose the min. weight of the matched terms
  8. ...?

So far about the rough shadow of some huge cloud above my head. Well you might say it sounds good but is hard to bring to live. I know, but I think we all could gain a lot from such a knowledge central, especially though it can use all places to look up Q&A. In fact this cloud sometimes looks like a small PQAN as something to be to Perlfaq as CPAN is to the modules.

But when I wake up the cloud is gone but it's shadow is still covering my thougts. :-)
So do You know of wheter this was implemented somewhere and how they did or do You have any idea where to start while trying to make such dream come true?

Have a nice day
All decision is left to your taste

Replies are listed 'Best First'.
RE: I have a dream ... (hits shown on preview)
by ybiC (Prior) on Nov 08, 2000 at 17:50 UTC
    Ya got my ++, little.

    Sounds a bit similar to something Russ and I discussed recently.   For any questions posted to SOPW or Q and A, the preview screen would include links to existing posts that might already answer the question.   You wouldn't necessarily require the questioning Monk to read them, just present them in case a newbie hadn't already searched.

    Is basically what Cisco does at their TAC website for contractual customers.   About half of my questions there are answered by the "hits" so don't get posted.
        cheers,
        Don
        striving for Perl Adept

RE: I have a dream ...
by brainpan (Monk) on Nov 09, 2000 at 06:16 UTC
    As I read your proposal I kept fluctuating between thinking "that's a great idea" and "people should check out as many different keywords in their search as possible anyway; this isn't the search engine's job". Since the purpose of most searches is to answer a question that you lack knowledge on, it's not unreasonable to suppose that you might not know / think of all of the keywords that might turn up the answer to your question, so your suggestion definitely does have merit.

    As to exactly what form this should take, I'd question whether anything more complex than a "smart synonym" type search / recommendation system would be worth the trouble. Even having the search mention that "related keywords to try would be..." would seem to settle into a fairly comfortable place when considering the law of diminishing returns. I think that implementing some of the changes mentioned on this node and adding this "smart synonym" feature to the existing search facilities would provide the best return for the time investment.

    But then again, what do I know?

      Well, if someone went through the trouble of adding the support for "smart synonyms" to the search why not then just use all this wonderful, portable Perl(the only kind vroom writes :) to then take the step further and do the check as one's posting in SOPW? It strikes me that getting the 'synonym' stuff right would be the hard part and parsing through the text of a new entry to stip out junk (non-keyword prepositions, proper nouns, monk names, etc.) would be the easy(ier) part.

      I'd also like to add that the way to display this info might be something like a sliding scale of matching. (this is oddly enough inspired by the 'what to expect' seiries of books that all parents in the monastery should be all too familiar with). The strongest matches could be listed as 'most possible', followed by 'probably relevant', 'possibly relevant' and 'shot in the dark'. And, if we're to consider Considering Super Search, maybe we could allow people to choose only strong matches, etc. etc. I would think that posts for SOPW would actually be parsed for the weakest matches, though.

      my $0.02

      "sometimes when you make a request for the head you don't
      want the big, fat body...don't you go snickering."
                                               -- Nathan Torkington UoP2K a.k.a gnat

        If I'm getting this (and I'm not sure that I am) there was a a module/article in the last but one or two TPJ about the perl frontend to somebody's extensive language dictionary/thesaurus (sp?) project.

        The idea was you could ask for, say, red monkey info and get baboon, great ape, etc. as the 'database' was able to know that each is a monkey. It was far better than that and, the perl part made it extendable (the linquist researchers who created the original didn't appear to want any help extending their db). If only I could remember a detail ...

        a

Log In?
Username:
Password:

What's my password?
Create A New User
Domain Nodelet?
Node Status?
node history
Node Type: perlmeditation [id://40515]
Approved by root
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others making s'mores by the fire in the courtyard of the Monastery: (4)
As of 2021-11-27 20:02 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    No recent polls found

    Notices?