8 thoughts on “Solar Magnitude Forum”

  1. Sounds perfect. User interface would be everything, and you seem to be describing tags for the content. Then tags for the member, and a rating system for how important various tags are to you.

    That’s a fine Masters project. The accompanying Bachelors project would be the import utility to bring in various popular web forum formats.

    What I like about it is that if you were ambitious (hi, Google) then you could say that the entire world should be on one interconnected system, and although I am NEVER going to have a tag for Mozambique or Orphan, I might have one for Injustice so a conversation far away among people I would never meet on a board I would never join might float into view.

  2. One of the key differences between what what you’re proposing and how it works today is the notion that the person who wants to participate in forums has to describe what they’re interested in. Right now as far as I know, nobody does this. The search engine has to guess at what you’re interested in based on results to specific questions.

    I think you might be able to accomplish a good percentage of what you are interested in just by having a person write a paragraph about their interests, and then using plain keyword matching from their interests to a list of forums where those keywords appear most frequently. No fancy indexing required.

    The larger question is how to formalize topics in a particular forum in a general way, and match a list of expressed interests from a user to the list of forum topics. You might have each forum stored with a data value that describes where it falls in some taxonomy. If the user’s preferences match any taxonomy element of a given forum, or any superclass of any taxonomy element, you could suggest that forum to that user. This would be a (possibly effective) simple way that doesn’t require some huge number of indexes in the
    database.

    I don’t think it’s a good idea to immediately jump to database indexes and then
    characterize the problem in terms of why those wouldn’t work. Using taxonomy references would only require the ability to join tables on one or two values (references to membership in some taxonomy). Maybe you could add some fancy sub/super class inferencing so the system would know that taxonomy reference “A.1.a” can match a reference to “A.1”

    I’m a bit unclear as to why google isn’t already doing this. They’re pretty good about finding group messages that apply to a given search. Why not just twist the search around a bit? Instead of searching for *POSTS* given a *query string*, why not just have the search result be the names of groups containing
    posts that often match an *interest profile*? Either way it’s just text search.

    I agree there are a lot of interesting issues surrounding characterizing the various axes of a particular interest…but the specific problem you’ve laid out (finding forums of interest) can probably be solved with novel application of ye olde Google with “dimensions” being faceted matches to plain text.

    Think also about the gmail metaphor of no folders, just search. This matching of users to communities might get really interesting if you abolished the communities as formal groupings, and had the groupings of posts arise as a consequence of their inferred topic area. You’d probably have to make special exceptions to keep off-topic threads in a particular grouping for social reasons, but such a system would make cross-posting unnecessary because posts would be visible to other groups by nature of their content.

  3. 1. Some of what you’re doing sounds like the personal training file of Bayesian spam filters.

    2. As far as “all the world is search,” at some point, this makes the forum no longer a virtual “place” where a “community” gathers, but simply a group of users performing their own personal searches, which may or may not interact. It seems like the difference between CBGB’s in it’s heyday and a group of people sitting in the same room wearing headphones connected to iPods: why not just stay home?

    Communities have regular participants and/or lurkers, and, setting aside some of the more psychosocial aspects that seemed en vogue circa 1994, I see utility from fora and usenet for the following reasons:
    2.1: When I post a question, I am likely to get a human answer…this ~may~ be reduced if everybody’s simply connected to the forum to the extent that each post matches their search preferences.

    2.2: When I post an answer, I may be motivated partly to increase my reputation. This is may be less likely if the “forum” is simply an unknown set of search preferences (would Philip see my reply if it doesn’t contain the right keywords?)

    2.3: The potential for slightly off-topic content from like-minded people is sometimes a feature, not a bug, in a well-designed forum. I definitely didn’t search for “solar magnitude” to get here, and I do not generally read stuff about information theory. So going back to the previous examples, a Beatles fan might be interested in hearing another Beatles fan recommend Mazzy Star, but would not generally search for “Mazzy Star.” Or a C172 owner who is lightly considering upgrading would probably read another C172 owner’s review of a Mooney.

  4. Yes, tagging is clearly the paradigm here, and the trick indeed is how to get people to tag their posts appropriately. You have to somehow identify what’s in it for the individual user so that “categorization” emerges naturally without burdening posters (Joshua Porter’s called this “The Del.icio.us Lesson”).

    Alternatively, you’re talking about organizing things using faceted metadata, another topic that people have been applying to the web for several years (mostly since Phillip’s book was published). See: Target, Tower Records, most travel sites, Amazon.com…anywhere you see multiple seperate navigation heirarchies like “refine by size/refine by price/refine by color/refine by brand”. The burden is still on the user to put the post in the right categories, but you can allow lots of combinations of categorization.

  5. I think that tags are part of the solution, but the meat of the problem is that the source material is read-only. It’s necessary to refine the categorization of source material after it’s submitted, and HTML/web servers don’t allow this to occur because everything becomes read-only.

    To do this properly, I believe it’s necessary to add an additional set of metadata to existing content which can, at minimum, be appended, if not outright read-write. This would allow the inclusion of additional knowledge from users as they review the source material, without the burden of making it writable (and thus subject to defacement, spam, etc).

    The collection of this third party info is the crux of the problem, and is probably itself worthy of a masters degree and/or open source software project or two. You’d have to tie reputation to the system to help cut down on gaming, and preserve some ownership. If someone can build reputation, this gives a positive incentive to contribute, and build a community.

    I hope this all makes sense, and adds value to the discussion.

    –Mike–

  6. Craigslist does something related to this: Their topical forums by default show discussion from all locations smooshed together, but you can narrow it down to just the location you used to access the forum.

  7. Why not just index the postings with Lucene (or an equally capable search engine) then use the “boosting” capability for the “brightness”? You’d probably have to take advantage of the scoring manipulation but the core code is there. Once it’s indexed, you’d get faceted browsing kinda for “free” too.

    E.g.
    for a canon home photographer in fairfax virginia:
    canon^3 virginia^3 fairfax^4

    The Solr[1] subproject is already appropriately name for being the engine behind your idea;)

    –tim

    [1] -http://lucene.apache.org/solr/

  8. The problem with associating a tag on a user basis is that most participants do not want to fill out long questionnaires about their interest in order to post.

Comments are closed.