Is Google Stuck in the Middle Ages?
July 6th, 2008
Google’s announcement this week that it will begin to index Flash content is a giant step forward for the search company which has historically left multimedia rich sites made in Flash out of its results. However, I often wonder if Google is stuck in the Middle Ages? The company is the king of text based indexing but has little capability when it comes to indexing multimedia.
Image search seems like a logical starting point. After all, a picture is worth a thousand words, right? To date, Google’s Image Search has relied on examining the text around an image and then determining what the image is about algorithmically. This works to some extent (I use Google’s Image Search all the time) but what about actually analyzing what’s in the image? There are several companies out there that already do this. Take Toronto based Idée. Give their TinEye Image Search Engine a JPEG and they’ll find other JPEGs that are similar to the original. I tried this out with a photo of the Eiffel Tower and they returned several images that contain the Eiffel Tower but weren’t the same as my original. Pretty cool! Their Visual Search Lab is also worth noting. Select a random photo and they’ll return other photos with the same colors. In both these examples, they’re actually analyzing the contents of the images.
Another image search technology Google should note is what Evernote has built. The service lets users collect links, photos, notes, and other multimedia and then search through it. If you upload a photo, Evernote will automatically add tags to the photo from text it recognizes in the photo, essentially indexing it. For example, uploading a photo of a Canon camera box with the Canon logo on it tags the photo with “Canon” automatically. The technology can read all sorts of fonts (the text doesn’t have to be boring) as well as handwriting unless you write like a doctor. Why isn’t Google investing in technology like this?
Google is even more behind when it comes to video search. Their video search is seriously lacking in sources (they only appear to index a couple of the top video sites) and their approach to video has been similar to how they index images by looking at the text around the image. There are a variety of more robust video search engines such as Truveo and Blinkx. However, none of the major ones appear to be taking advantage of speech to text engines, object recognition, facial recognition and other technologies that actually examine the content of the video. Companies like Everyzing, Viewdle, and Digitalsmiths all offer these technologies.
Maybe Google is waiting for the technology to mature or for another large company to take the first step but if they don’t move beyond just examining text, another company could take their place to dominate a growing multimedia search market.
Sphere: Related Content