Author Topic: Auto-Tagging?  (Read 1133 times)

Offline The Littlest Leaf Dragon

  • Alis Volat Propriis
  • Teacher's Guild
  • Evaluator
  • *****
  • Posts: 2350
Auto-Tagging?
« on: August 13, 2010, 06:43:23 PM »
How difficult do you think making an auto-tagging feature would be?  I mean, the code wouldn't be that difficult to do, of course, but you would have to pass it through a filter so only relevant words popped up, and then you'd have to integrate it with the forum somehow, either via JS or PHP.  Do you think it would be worth it to make it a public suggestion after all of the effort?

Offline mu

  • Ishmael
  • Technical Guild
  • Evaluator
  • *****
  • Posts: 4584
Re: Auto-Tagging?
« Reply #1 on: August 13, 2010, 11:37:11 PM »
What exactly do you mean?  Something that keeps a sort of frequency count of words, perhaps removing common words?

Other than removing common words (it, the a, etc..) I'm not what sort of filter there could be..

Unless people could choose the words to filter for, and have dreams containing such tags be auto-tagged..

Any sort of word search, filtered or not should not be too difficult.  I haven't looked much at the way Raklet implemented the whole thing; how easy the rest is depends almost entirely on that.

Offline The Littlest Leaf Dragon

  • Alis Volat Propriis
  • Teacher's Guild
  • Evaluator
  • *****
  • Posts: 2350
Re: Auto-Tagging?
« Reply #2 on: August 14, 2010, 12:26:17 AM »
Well, you'd have to filter out the common words, like you said, but you would also have to filter out some of the other words, since you'd be going for dream signs.  Consider my dream below, for example.

The earliest I remember in this dream, my grandma had some glass contraption with a jar inside containing a thermomiter suspended by a pipe in her bathroom.  I had undone it, and was trying to fix it.  I finally screwed it back onto the fixture the right way, and looking at it I thought there might be spiders on it, but they were only screws. *skip*  Now I am in my old bedroom, sitting on my bed, and there were two guys that were slightly older than me at my computer desk, one sitting down, who I didn't clearly see, and one standing up, who was slightly taller than I am, wearing a blue and red button down shirt with jeans, and is skinny with black hair.  The one standing up was explaining something to us, I thought it was about HIV when I woke up, but it had something to do with a spider bite on the neck. I remember calling him over (at this point I know he's my older brother) and I don't clearly remember what happened next, but at the end of the dream, he was laying on the bed (for a shirt he only had a white undershirt on now) and I was laying with my head on his chest as we fell asleep.


Here is an analysis with the most common words in it, removing stop words, of course.

Word   Occurrences   Frequency   Rank
remember   3   3.1%   1
shirt   2   2.1%   2
slightly   2   2.1%   2
bed   2   2.1%   2
thought   2   2.1%   2
sitting   2   2.1%   2
something   2   2.1%   2
now   2   2.1%   2
laying   2   2.1%   2
older   2   2.1%   2


So, assuming we took out now, we would be left with shirt, bed, sitting, and laying as the best words, the rest being useless.  In fact, anything ending in ly would probably have to be removed.
Source

Offline mu

  • Ishmael
  • Technical Guild
  • Evaluator
  • *****
  • Posts: 4584
Re: Auto-Tagging?
« Reply #3 on: August 14, 2010, 12:45:25 AM »
I wonder if there isn't some open source app or library that does this..  Intelligently analyzing language in this way isn't exactly trivial; to come up with consistently useful tags would be challenging.  For example, frequency may not indicate at all what the subject really is; something may be referred to again and again but with pronouns..  Text matching, no matter how clever will never be able to do this well IMO.

As for how this could work; I could see having a "guess tags" button near where they're entered manually, so that they're the same as any other tags, and can be removed or added to.

Offline AspirationRealized

  • Aspirant Oneironaut
  • Chat Moderator
  • Evaluator
  • *****
  • Posts: 1824
  • The Cat's Pajamas
Re: Auto-Tagging?
« Reply #4 on: August 14, 2010, 12:58:50 AM »
Yeah they should display as suggestions, not automatically.
Drow, drow, drow your canoe
 The stream provides flotation
 Hysterically, hysterically, hysterically, hysterically
 Existence is hallucination

...have you ever met anyone who actually changed?

Offline Sean999

  • Creator's Guild
  • Evaluator
  • *****
  • Posts: 974
  • Only in your dreams...
    • The Sound of My Universe
Re: Auto-Tagging?
« Reply #5 on: August 14, 2010, 01:11:48 AM »
Hmmm, I'm not sure about how others tag their dreams, but when I used to, I'd just have one tag (per post) for whatever needed tagging, regardless of its frequency. So, at least for me and maybe others, pronouns aren't an issue as long as the subject is mentioned at least once.

I think suggestions would be nice. Maybe you could give the user options to filter out nouns, or verbs, or adjectives, or whatever. Not sure how that would be done (maybe by screening the words against separate libraries for each group?).

Something I just thought of, is "related words". For example, if I had a dream with my brother in it, I'd probably tag it "brother", but also "family", even though I don't necessarily have the word "family" anywhere in my post.

Just stuff to think about.