h a l f b a k e r y
Tastes richer, less filling.
idea:
add, search, annotate, link, view, overview, recent, by name, best, random
meta:
news, help, about, links, report a problem
account:
Browse anonymously,
or get an account
and write.
Login
Create account.
|
|
|
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
I've long had an irk about online forums and the similar, in that they tend to emphasise the current and recent activity among participants, but pretend that they focus on a specific topic (usually in the form of the thread 'title') instead.
People will tend to talk about whatever's on their mind
at the time, and have to be shepherded back into agreement with the 'topic'. Of course, this is a constraint, some agree with, some might not.
What might be an idea is that different threads get started with no topic, no focus, and just using the magic of computers and programming, 'derive' a topic based on what it thinks the discussion is about (and this should be dynamic, and perhaps fairly 'deep' as well as having a simple surface 'label' form).
It should also be able to describe the thread by what it might know about the participants, and what it might know about related topics not yet mentioned but would otherwise form part of the 'set'.
It should also be able to describe the time that the main activity on the thread took place and use that as an influence in the topic description (perhaps a meme is going round at the time, or perhaps lots of people are currently under the impression that a thing is so, whereas it's not, but this doesn't become apparent until later or earlier) (or something).
Total Non-Sequitur
Total_20Non_20Sequitur
Reminded me of this magazine... [Jinbish, Oct 23 2007]
Text Summariser
http://search.iiit....ummarizer/index.cgi
A bit and miss mut could be tweaked to provide dynamic titles for threads. [boysparks, Oct 23 2007]
Not location specific, no rules and pants are optional - Could this thread be heaven?
http://www.ilxor.co...d=40&threadid=54269
"This is the thread without purpose, focus or direction" [calum, Oct 25 2007]
Annotation:
|
| |
Do all these annotations have to be about your idea? I went canoeing at the weekend. |
|
| |
Wouldn't it scare the sparrows? (the sheepdog, not the canoe) (or the weekend) |
|
| |
i thought it said "topless forums." i was scared to come in. |
|
| |
//i thought it said "topless forums."//
So did I but That's the only reason I came in |
|
| |
A kind of virtual dinner table chitter chatter. Nice idea. |
|
| |
//'derive' a topic based on what it thinks the discussion is about// dynamically updating. |
|
| |
This could probably use something similar to a text summarising system (set to output one sentence) - see link. Quality would be variable. |
|
| |
The recent activity among conversation participants is mostly pretty boring. No matter how much computer linguistics you throw at it, finding out that X just made tea and Y is playing with her cat won't turn those into long-term relevant statements. Are you sure that you're not trying to inject meaning into something that very fundamentally lacks it? |
|
| |
You may now change the title to "Topless Meaning Activity <undefined> Sheepdog Topic #REF!" |
|
| |
//Are you sure that you're not trying to
inject meaning into something that very
fundamentally lacks it?// - also, no one
likes judgemental computers. If 90% of the
time the forum topic turned out to be
"Dull, pointless trivia, again" then the
forum participants would start to get a bit
irritated. |
|
| |
This would derive meanings from superficially meaningless chatter in a manner akin to Freudiam dream analysis. I am sure some discussion participants will be horrified to learn of the subtexts of their discussion, but there will be no arguing with cold machine logic. |
|
| |
hippo, - you'd think so, but last time I looked, facebook was still popular. |
|
| |
jutta, - to be honest I wasn't anticipating that topics would comprise of nothing but inanity, that somewhere in the soup there'd be some actual question/answer intercourse, and perhaps the formation of useful knowledge. |
|
| |
My irk was more along the lines of, for example, flickr groups, in which I spend quite a bit of time, developing into the same old patterns I see elsewhere: threads form under topic titles that vary from misleadingly useless to fairly succinct to stream of consciousness. Within the thread some form of topic evolution will occur, possibly reaching an absolute resolution of knowledge. Then at some point, the topic - perhaps past its peak by then - drifts or accumulates bolted-on aspects, and before long it sinks from sight. |
|
| |
A few weeks or months later, a near-identical topic forms, the knowledge plateau is established by various means - reinventing the wheel with new participants, quick interjection of knowledge by linking or directly relaying it thanks to available participants in the previous thread, or it just idles for a while until someone rants about how many times this is in the archives if people could be bothered to search. |
|
| |
This happens over and over and over. It's because of the temporal nature of thread formation that I keep banging on about. But this idea's solution is to accept that people have a novelty thirst, and why not keep all threads alive by unshackling the boundaries of threads. Experiment with no topics - or at least, no topic in the title. But then, how would one identify a thread? |
|
| |
I imagine that as you type, keywords from your text - perhaps extended via semantic fields - spawn queries that search past conversations for similar subjects. These conversations then bob up on the screen like memories in one's mind. A conversational Deja Vu Home where new comments get tacked on to the revolving donut of past dialogues. (Mmmh, revolving donut.) |
|
| |
Two obvious problems with that: (a) people can't read for shit, (b) half of the time, I'm not talking to you because I want information; I'm talking to you because I want you to like me. |
|
| |
Maybe an easier to implement, shallower variation is to allow forum users to tag a thread with keywords, or better still phrases, and then crucially alter the 'relevance rating' of that tag as the conversation evolves. |
|
| |
The tags would appear at the top, with relevance indicated by font size (as used elsewhere ad nauseum). |
|
| |
If a user feels that a new tag is needed to reflect the thread's current topic they could add one. Other users could then click to strengthen or weaken that tag's relevance as they saw fit (through the use of cute little +/- buttons next to it). |
|
| |
Maybe the perceived tag-relevance could be temporally recorded, so as you scroll through the annotations in date order the sizes of the tags at the top of the screen alter, alongside new ones appearing and others shrinking out of existence. |
|
| |
Apologies if that kind of thing already exists. |
|
| |
If it doesn't, maybe the +/- aspect could be meshed with the beta version of this idea to allow users to provide training feedback for the computer-generated topic tags. |
|
| |
//I'm talking to you because I want you to like me.// |
|
| |
We like you, [jutta], we like you. And this observation of yours is not fatal to this idea. What it implies is that the main semantic payload of those phatic postings would consist of a contribution to our knowledge of the relationships between posters. |
|
| |
Technically, it could work like this: the site would maintain a model of 'relationship space' based on an extension of graph theory. Whenever someone posted, the site would look for things in the post which connect the poster to other posters (such as common references or mannerisms), and feed that data into a number indicating how much poster X thinks of poster Y, that number being associated with an edge on the graph. (This 'relationship space as graph theory' thing is an idea I've published elsewhere). |
|
| |
You'd need to combine this information with a bit of lurkometry - maybe cookie-based lurkometry. |
|
| |
Anyhow, all this could get the conversation tagged, not as "dull pointless trivia again", but as "[X] is on the flirt again", or "clique alert". |
|
| |
Is that what blogatars do? That's not the impression I got from blogatar.com |
|
| |
...wanna talk about cameras? |
|
| |
Ian *always* wants to talk about cameras. When he's drunk he starts slrring. |
|
| |
he's great with beer too, & sex & lots of other stuff - he multi-tasks, believe me! |
|
| |
actually, could I do a poll here? does anyone believe me? |
|
| |
In general, or about Ian? |
|
| |
Well, I'm not in a position to confirm that sex with Ian was "great", but I can confirm his enthusiasm to talk about photography and beer. |
|
| |
//they tend to emphasise the current and recent activity among participants// |
|
| |
the.jxc has changed the topic to: "Wibble". |
|
| |
Thus, the title of this invention should be changed to "Favorite Animals Temporal Beer Relationship Space". |
|
| |
What were you saying again? |
|
| |
I forget and it all came out wrong anyway! |
|
| |
//it all came out wrong anyway// |
|
| |
But I thought you said he was good at... oh, never mind. |
|
| |
With more sophistication, you have invented a generalizing AI... |
|
| |
Rather call it "Collatz title generation", as po alluded to, most conversation, from any given starting point, will degenerate to Politics, Relegion, Sex. Akin to the 4, 2, 1, sequence. Will save the programmers some bother. |
|
| |