Comment not quite 'mislabeling' -- 'inaccurately' labeling (Score 1) 389
While I'm down with having humans drive databases that computers could never drive, doesn't anybody think that this labeling technique is lame-ifying google images? When I played, the first image I got was of a computer and a scientific instrument. I have no idea what this scientific instrument is, so I type 'computer'. That just feels backwards.
Then I got the backside of a dvd whose title I couldn't make out. My question become "why should I be labeling these things when I don't have any insight."
The question of insight or prior knowledge is important with regards to context. Lets say I'm posting a preview image for a Wordpress Skin. The classic google image tags associated with this are "Wordpress Theme," which contextually fits if you're a user searching for wordpress themes.
However, if you look at an image of a Wordpress Theme, you won't label it either Wordpress or theme, you'll label it "blog", but if you're searching for images of "blogs," wouldn't you want to filter out "lorum ipsum" entries?
If I'm looking for images with the term "scientific instrument," I _expect_ to find something vague and cliparty, whereas if I look for images labeled "electron microscope," I expect very specific results.
My point being that this tool is muddying the distinctions between highly specific imagery and a vague third-party notion of what the image may be about.
What do you all think?
Then I got the backside of a dvd whose title I couldn't make out. My question become "why should I be labeling these things when I don't have any insight."
The question of insight or prior knowledge is important with regards to context. Lets say I'm posting a preview image for a Wordpress Skin. The classic google image tags associated with this are "Wordpress Theme," which contextually fits if you're a user searching for wordpress themes.
However, if you look at an image of a Wordpress Theme, you won't label it either Wordpress or theme, you'll label it "blog", but if you're searching for images of "blogs," wouldn't you want to filter out "lorum ipsum" entries?
If I'm looking for images with the term "scientific instrument," I _expect_ to find something vague and cliparty, whereas if I look for images labeled "electron microscope," I expect very specific results.
My point being that this tool is muddying the distinctions between highly specific imagery and a vague third-party notion of what the image may be about.
What do you all think?
