free html hit counter Google Image Disambiguator - John Battelle's Search Blog

Google Image Disambiguator

By - September 03, 2006


Why does Google hate the word tag?

Google Image Labeler is a good idea but a terrible name – it sounds like a product Dymo might have created. Google has licensed “The ESP Game“, invented by Luis von Ahn, and is harnessing collective intelligence to tag images in its image database. This is just an experiment, of course. But it’s a clever one – -if a critical mass of images are tagged, Google will have solved a very intractable problem for itself.

I just wish Google would use the terminology the rest of the web has already settled upon. It’s not a label. It’s a tag. “Tag” means something – an intentional attribute given to an object on the web. That’s what we are doing here. How about we help Google come up with a new name?

If tag is off limits because of competitive issues (after all, Yahoo owns Flickr), I vote – with only a slight portion of my tongue in cheek – for “Google Image Disambiguator.”

Or maybe, Google Metadata Maximizer.

Or…what do you guys think?

(Also, you can earn points for tagging on GID. But what can you do with the points?!)

Related Posts Plugin for WordPress, Blogger...

19 thoughts on “Google Image Disambiguator

  1. At first thought, I was thinking like you. But after using it and saw the same image coming back with the “off limits” tags word I have already given, it seem this game can give some sense to image.

    If they can make algo to produce good text results, they certainly can make algo to label the image with the better tags no ? I put my point of view on that (in french) here !

  2. There used to be a popular game show in the UK where the presenter would ask the audience: “What do points make?” the chorused answer coming back: “Prizes!”

    You could go one step further and get Google to parody itself: Goodle Image Tracker (GIT) but that might be seen as tasteless by some. πŸ™‚

  3. Tudor says:

    Google is doing a smart move, with that . Why? Because it doesn’t disappear in the big sea of tags. No, wo don’t tag we label.. we are google, we are unique πŸ™‚

    Very smart.

  4. Alex Iskold says:

    I actually played the ‘game’ and I have to agree with John. First, there is no point in inventing new names. Labels are tags end of story. Using standards is good because it makes people immediately recognize what this product is doing.

    There is a bigger problem, however. People are not going to be doing this for fun, because this is not that much fun. There needs to be a feedback loop for people to get excited. Flickr approach is better because everyone is tagging their own content.


  5. Raffy says:

    We just don’t know how Google is going to implement these “labels.” The obvious choice is to attach them as invisible metadata to each image and return that image when a user searches for a matching keyword. However, it could extend beyond that – for instance if they chose to use synonyms from Google Sets in searches.

    Tagging implies a very specific implementation of “an intentional attribute given to an object on the web.” When there’s tagging, I picture tag clouds and isolating filters where I can find all objects with the same tag. Google may not be going in that direction with image search.

  6. Narendra says:

    I think it is completely off focus. Google has not yet been able to index the 420 million photos that live on Webshots and have a wealth of metadata associated with them. They need to focus on improving their crawling ability and be able to play with large existing systems like Webshots and Flickr to improve relevance instead of forays into odd and ambiguous user annotation.

  7. fantomaster says:

    1. Full agreement on the “tags” contention – they’re not labels, so why call them such? Isn’t that the kind of we-are-so-special-we-will-do-what-we-like-never-mind-the-standards attitude that once made Microsoft one of the best hated companies on earth within the Web techie community?

    2. What’s next: spam tags on “Texas holdem poker”, “viagra”, “weight loss pills” and the like galore? It’s happening on the social bookmarking sites already so what will this spawn if not yet another spammers conduit?

    3. Taking the easy way out by letting users do the work for them, attempting to lure them with a puerile “grab your points” pseudo-challenge without any real community character involved rather than hire qualified staff may well serve to achieve wildly sub optimal results provided by a huge bunch of the essentially clueless and/or those with too much time on their hands. Not to forget hordes of practical jokers.

    It’s the old story: if you pay virtual peanuts, you get cyber monkeys.

  8. SutroStyle says:

    Google hates the word “tag” because initially there was no web search, a yahoo directory of web links was a collection of tags, or keywords (like AOL keywords). Google search was intended to render tags or keywords obsolete. Now they have a hard feeling about a trend of reverting to tags again: it’s agains their company “DNA”.

  9. karaterobot says:

    I think their use of the word “label” is correct. Tags are a kind of label which are wholly uncontrolled and idiosyncratic, but the point of Image Labeler is (presumably) to find consensus in the labels given to images, so they’re really not idiosyncratic at all.

  10. JG says:

    Tudor writes: No, wo don’t tag we label.. we are google, we are unique πŸ™‚ Very smart.

    Fantomaster writes: Full agreement on the “tags” contention – they’re not labels, so why call them such? Isn’t that the kind of we-are-so-special-we-will-do-what-we-like-never-mind-the-standards attitude that once made Microsoft one of the best hated companies on earth within the Web techie community?

    Look, I do not work at Google, so my guess is as good as any, I suppose. But I have to disagree with most of you who have commented here. The overwhelming consensus seems to be that Google is doing something new by using the word “label” instead of “tag”. How wrong that viewpoint is!

    If you look at the field of machine learning, which has many of its roots in AI, you will see that there has long been classification of machine learning algorithms into two approaches: “supervised” and “unsupervised”. The difference between the two comes down to whether or not there is labeled training data on which an algorithm can base its inferences.

    Machine learning itself has been around longer than Google has existed. The word “label” to mean some sort of keyword associated with an object has been used in the machine learning community for quite some time, and definitely predates all this Web 2.0 “tag” hoo-hah.

    Now, if you look at Google, you will realize that they hire a whole lotta machine learning PhDs. It is therefore only natural that folks with that experience would choose the word with both historical precedent as well as current active usage in the literature.

    Maybe in the Web 2.0 world, “tag” predates “label”. But that is one of the big problems I have with Web 2.0..they are just inventing the wheel all over again, and have no clue as to what went on before. Web 2.0 does not predate Machine Learning.

    For that matter, don’t even get me started on the word “folksonomy”. Thirty years ago there were debates in the information retrieval community about manual vs. automatic indexing (i.e. searching documents by using either human labels (“tags”) or full-text content). And within the manual indexing area there were debates about constrained versus unconstrained vocabularies. A constrained vocabulary is a taxonomy. Wanna guess what an unconstrained vocabulary is?

    “Tags” and “folksonomies” are nothing more than unconstrained vocabulary, manual indexing.

    So, who really knows why Google chose to use the word “label” instead of “tag”. But by so doing they are actually returning to historical roots, rather than jumping on some bandwagon.

  11. dumbfounder says:

    The general public understands the term label but I have found they don’t understand tagging right away. But I still think ego is a factor with this nomenclature choice. Tagging is what it is, just get onboard already.

  12. joe says:

    Likely the PM is one of those stuffy types who uses words like “Hoo-Haw” to describe nascent technology she’s failed to comprehend ..

  13. Matthew Hurst says:

    IMHO, tags are a mess – perhaps Google shares that opinion. Majority doesn’t mean right.

  14. Google have historically done well at improving on what others are already doing. To try and change it could be a bad move for them (they’ve made quite a few recently – GoogleVideo anyone?)

  15. John says:

    “Tag” was used long before Flickr… Try Adobe Photoshop Album. That functionality now is in Adobe Photoshop Elements.

  16. JG says:

    Joe: Someone who uses the word “hoo-hah” is a “stuffy type”? Oy vey.. go back and study your Yiddish.

    John: And when did Photoshop Album come out? 2003? The word “label”, used in the machine learning community to mean an annotation for an object, such as an image, still predates Adobe by at least 25 years, if not more. Adobe, much less Flickr, did not even exist in the mid-70s.

    Face it, “tag” is a recent re-invention. “Label” is the historically correct term. Google has it right on this one.

  17. neuroxik says:

    I think you guys are too focused on the naming of the product than its actual purpose. Who cares if it’s tags or labels? As long as the underlying engine algorithm calculating the input does a great job.

    I read someone mentionning spamming. I don’t think that would be a problem. Google most probably uses a cookie to verify if different users post the same content, (Google uses cookies everywhere), so if only one person spams “Texas Holdem Poker”, then it will not be considered.

    At first I was skeptical too. For example, I see a picture of Ron Jeremy, I write Ron Jeremy as a “label”, but my label does not match off-limits’ tags such as “man”, therefore no points are earned. But remember, these images will be shot at more people playing the game, and if someone else mentions the label “Ron Jeremy”, then that image will have more value attached to that name than to the more commonly used “man”, because Google is smart. If the image has been exposed to 60 people and 50 labeled it “man” (among other labels), if 5 out of 50 said “Ron Jeremy”, then “Ron Jeremy” will have more power than “man”. This makes no sense? It actually does– Google knows what are the common words (car, tree, sky, river) and it will focus and specific names and words that are repeated.

  18. Google trying some human intervention, finally.

  19. Google Image Disambiguator or Google Metadata Maximizer or Google Systems Integrator Spacial or Google Information Systems Data Hal or Goggle Eyed Google Operators Vacationing is not the intention of Google hopefully competitively per se. Google notices, collects user’s information investigations for purposes relevant.(No let’s not get into Homeland Security Issues) Clips, blurbs, blogs, profiles etc. accumulating at rates byte wised “Tags” are understandably smart by Google. “Tags” created by users reflect true said knowledge seeking as computer programs, computer engineers reticently proliferate and promote as does harnessing refinement in the computer. “Tags” shouldn’t be deleted from Google but promoted as this what users want to see. Competition ex. Yahoo, MSN, & networks not as giant ed need not de-data incoming users preferences as coursed for long time “dots” or as nothing but processed material will evolve conforming to idolatry of mega monopolized competitor based on financial, marketing terms. Let’s have Google Metadata Maximizer totally disregard paid websites in when users refine vocabulary. Not all users are financially credit scored in the 800’s or credit scored at all but these users globally need the information that they seek or why would they go to Google? “Google Free” or “Google Let Freedom Ring” are alternative terms divulging disheartening notices from Google when good “Profiles,” of citizens are cached, hid, or erased. Google should re-structure time limits of Profiles, articles, etc. Erasing and deleting and I have my 2 good erasers when I write my notes too fat, etc. or decide to write a whole rest note on manuscript paper which I need a new book, nicer pencils, a new memory card(oh I can remember everything…yes) Blank, blank…or there’s Google! Peace!