in reply to Anyone doing anything about questions/answers at PM?

I've made suggestions at least a few times about solving this problem but no monks ever seemed interested. My point has been that:
1) the same questions always get asked,
2) it is hard to view recent threads once scrolled off a page, and
3) Considering the monastery metaphor, it would be very useful to develop a tool to focus the accumulation of the body of knowledge being collected, probably using a collaborative tool, making things more like a wiki(pedia) and less like a mailing list archive. (Granted PM is not that bad but it is more opaque than necessary).

Personally I am happiest when I can find well organized documentation (the online user-annotated mysql manual is nice, and the Mac developer documentation was too, haven't seen OS X stuff yet). Perldoc.com is kind of nice. Maybe someone has other better examples of good online knowledge bases.

How about this?

We need a collaborative tool built into PM that allows us to input which leaves of a tree a given thread/node should be linked to. Viewing a thread/node will show which leaves it has been registered to, potentially with voting involved.

I've thought of building something myself and hosting it for use by monks but it would probably be best if supported by a nodelet in PM.

So the tree structure would be editable and annotated to allow addition of more refined topic areas, plus adding an unlimited number of well-written faqs, articles, code snippets, and link lists, possibly each could include at the end a section including entries automatically generated by scanning nearby leaves. A map of crosslinks indicating related topic branches/leaves and entries in perldoc or PM library might be good too.

An ontology of topics and where they fit in a tree might be something to consider, as a guide to topics to create when needed, but I am thinking it would be more useful to make this work like a faceted metadata engine (look at for example in Flamenco which works like this. The point here being that you never see an empty topic in the tree, and the tree is presented showing the number of entries per branch or leaf. And that you may use it in connection with a keyword search, but the main idea of faceted metadata engines is that you never end up with "no matches" or "a zillion matches", and you are not bound to navigation up or down in an hierarchy but you can always see where all information is in relation to where you are in the knowledge base (user-centric navigation). Also I think it might be useful to have a history of things you have looked up and maybe show all those for today in a split screen.

Editors could move bits around and prune, perhaps making better use of their time. Recommendations can be made for writeups on topics.

Newbies would then read the library and know right away where to go and drill down. You could publish a selection in book form to make money (books on how to remane sane in a company, what strategies to use, poetry and meditations, etc.) You could start building it right away with some things loaded into it marked for consideration/editing, and an automatic search listing could show you other nodes to consider adding, with the idea that the knowledge base could hold every node but there should be a well-defined division for a given leaf/branch between highly edited answers (like a wikipedia entry), annotations, and then all the other nodes that could match the topic. Then you could browse the KB with or without the distraction of everything but the kitchen sink. Maybe also sort annotations chronologically or by XP? Finally a good use for XP?

Well I hope this helps. I just thought this up now since my original idea (for the past year or two) has been to start with an empty wikipedia like KB tree to which monks would add writeups, and not try to make it a reflection of the massive amount of data in all of the PM database. The above suggestion though includes more community and maybe more interesting.

Finally I think it would be very interesting if, unlike PM, an API (Soap?) was provided so that people could make on their own servers tools to work with the knowledge base and also with the PM database, to experiment with tools for annotation, searching or navigation, and display. This would greatly accelerate development of the PM knowledge base and make it practical for the community to get involved where currently it seems you have to be a God working on Everything.

The brief description I provided above on faceted metadata engines outlines a large project that could best be done by people not on the PM server I think, and could maybe even take advantage of software not developed by a monk.

This idea I've had is part of a concept I've been working out for a long time on how to focus information contributed in an ad hoc fashion to solve a problem (the initial case being how to assemble goods to help people in an earthquake when Internet is the only communications medium available). It has to be easy for lots of people to add information, and actively work toward gathering and organizing information, keeping people involved in single threads for a long time, which seems to oppose the web's "constantly surfing elsewhere" paradigm. I suppose this concept embodies parts of metadata engines, collaborative definition of an ontology, annotation and wiki-like knowledge bases. I have no idea if it will really work though if it does, it will be pretty quick I'd imagine. Currently I'm working on a wxPerl based application that may eventually be able to do some of this resource focussing in a distributed fashion. Well if I ever get anywhere on these things that might be neat but please consider some subset of the above suggestion as potential points for discussion.

Interested to hear your ideas!

  • Comment on Re: Anyone doing anything about questions/answers at PM?

Replies are listed 'Best First'.
Re^2: Anyone doing anything about questions/answers at PM?
by BioGeek (Hermit) on May 31, 2005 at 16:04 UTC
    Before diving headfirst into a tree-based ontology, please (re)read this essay by Clay Shirky: Ontology is overrated, in which he makes the case that ontological classificationis is a bad strategy when your domain has a:

    • Large corpus
    • No formal categories
    • Unstable entities
    • Unrestricted entities
    • No clear edges

    and your participants are:

    • Uncoordinated users
    • Amateur users
    • Naive catalogers
    • No Authority
    In such cases a del.icio.us-like tagging system would make more sense, he argues.
      Thank you very much for your reply. Yes, I think your point is correct. I think I was arguing to use a tag-like (well I said faceted metadata) system instead of an ontology, but for this specific project I was more interested in having an ad hoc approach where people could add categories if nothing appropriate existed, and people could edit categories in a wiki-like manner. Well I didn't have that totally worked out for a large group, which is what I wanted to support, and certainly the del.icio.us way is very interesting.

      I am curious about how useful tagging would be for both my original system (intended to focus solutions to sub-problems for use in helping earthquake survivors) and the PM knowledge base which certainly seems to have some solid categories.

      But as I think about it now it does seem that the key here is that a plurality of people will be involved in an editing or knowledge focusing process, and since they share the same vision of focusing knowledge and using perl, then the tags they select will be relatively focused as well. In which case a separate set of "narrative threads" would then need to be added by people, for example in PM they might indicate a favored order in which to view a set of nodes for use as an introductory text. In my ad hoc disaster aid system I think it also needs an engine or a human editor who drives forward the overarching requirement that we find a concrete answer to a problem within a given time frame (like who will bring what, where, for someone to drive to the disaster site). There can be all kinds of side discussions going on but the main thing I wanted was to solve a problem.

      I had a vision of a very simple user interface with a small text entry box and a pull-down menu with subject names to select from, or you could add your own subject (or tag I guess). If you start up a new project to solve a different problem, you would have a different set of starting tags. I thought I would have a person periodically reorganizing things and reducing the number of tags as much as possible, or adding new ones (developing themes) to keep everybody headed toward resolution of the problem in a finite amount of time. This is not quite as important for PM perhaps.

      Well, thank you for a really interesting note, it has gotten me thinking in new directions since I had not realized everything del.icio.us could do until now, and it seems to lead the way toward better architectures.

      In the PM KB, it would seem then that you could derive a lot of information (metadata) from nodes (if you had access to the database) like who posted what when, how many votes the thread and each answering node got, and linked nodes and sites, which could provide a lot of initial tagging information.

      I think the most interesting thing here is that user XP, which has been said to be useless, and votes per node, may in fact be the most useful tools we have to determine who our most influential editors are and what threads and nodes have served monks the best. Perhaps we already have sorting by votes built in to Super Search or something else? Best nodes this month seems also to be a good source of informaiton.

      To get farther on this I think we need a corpus to look at, in other words start messing with WWW::Mechanize or something else to examine PM in these terms, or obtain readonly access to the database.

      So thanks again and looking forward to hearing more from you if you have time. Yes, I agree with you about ontologies but would like to experiment with adding structure (albeit changing and minimal) to tags, and to add tags that have meaning in terms of an overall objective (i.e. solving a problem, or explaining perl and good solutions well). Great!