The Social Media Oracle

(click for larger view)

Broadcast U

We are swimming, some would say drowning, in a sea of personal information on the web. Social media sites like Facebook, Ustream, and Twitter, make it easy for us to broadcast our lives online. Most companies with an online presence have been focused on SEO, or search engine optimization, in order to have a high page ranking by the likes of Google, Bing, etc. SMO, or social media organization, is the next step in harnessing the power from the shear amounts of information created daily on social media platforms. Such information, if identified and categorized correctly, would be a boon for businesses.

or•a•cle [awr-uh-kuhl] - any person or thing serving as an agency of divine communication

The difficulty for businesses is being able to filter relevant information from the noise. To break the codes, we need software to act as an oracle for  social media. The software would need to identify, contextually understand, and find the relationships between the mass amounts of text, pictures, and videos, found on blogs, Twitter, Facebook, YouTube, et al.

What did you say?
Self learning programs will be required to connect the dots from the macro to the micro. Software to find one's social media presence and gather all of the data from each platform is relatively easy. Taking the available information and extracting relevant patterns or information is finding the needle in the haystack - complicated but not impossible. Could we logically assume that two users who post the same picture of a sailboat on their Flickr accounts have similar interest? Not if you also knew one user was a Facebook Fan of the America's Cup while the other users profile expressed an interest in model building! Figuring out this relational data between the vast amount of information is akin to Google's search algorithm. Instead of finding the best search results from entered text, the Social Media Organization needs to find how the known information relates to each other.

A Picture is worth...?
Pictures are worth a thousand words, only if they speak to viewers. The problem is being able to logically infer the context of a picture from the billions on the Web. Simply put is that a picture of your dog, your neighbors dog, your favorite dog breed, or just a cute picture you saved? Deciphering the context of a picture is far more difficult than other data because pixels cannot be categorized as easily as words.

(note: picture is not the 2008 Westminster BIS beagle, but ain't it cute?)

• Meta data, or keywords, can be associated with pictures. Keywords for the previous example could have included: beagle, Fred's dog, 3 years old. Keywords can provide critical information that cannot be readily determined from the picture itself. The issue with 
• Geotagging, or knowing where the picture was taken, would provide additional contextual information. However, much like meta data, it is a burden on the end user since most consumer cameras do not automatically capture and "tag" the photographs. Most shutterbugs have to use cumbersome software to add location information hours and days after the picture is taken.
• Reverse image lookups like TinEye search the Web to find all known locations of the same image. Although the technology is probably geared toward copyright management, the site has the power to provide important relational data about a picture. By performing comparisons searches of meta data and keywords of the websites that are linking or displaying the image, a the intent of the picture becomes clearer.

A combination of these search techniques would offer a high level of accuracy in determining the relational information of any picture. Marketers could then tell that the dog in the aforementioned example was the winner of the Westminster Kennel Club dog show and not your own.

You, the movie star!
Videos present the biggest technical challenge. There is little to no meta data or geographically accurate information associated with online videos. Post any copyrighted material on YouTube and the copyright police politely inform you of your infraction. YouTube can accurately compare the "digital signature" of the data and compare it to a library of known, copyrighted material, but YouTube largely doesn't have a clue as to what the video is about. Interestingly, this is one of the main obstacles to attracting advertisers to YouTube as brands want full placement control.

Unlike text and photos, videos is not as critical to Internet driven 1-to-1 print material. Understanding all data points would help anyone interested in correlating one's digital persona with their real life.

Shoutback: These are just beginning points, what else should be considered?

photos by: blprnt_van, bredgur