Analyze This: The IBM–USC Annenberg Lab Social Media Analytics Experiment
Jimmy Kimmel pulled off an incredible prank during last Sunday night's Primetime Emmy award show. He got comedian Tracy Morgan to lie flat onstage and asked his audience members to tweet “OMG Tracy Morgan just passed out onstage at the #Emmys. Turn on ABC now.”
There was something fascinating about this prank. The notion that audience members are primarily glued to their television sets and would occasionally drift off to other media such as social networks is now dead. Jimmy's prank recognizes that a majority of his target audience members reside on social media and they needed to be veered off to watch his show. He had to draw their attention in a way that would compel them to “come back” to watching his show on TV. That is the new reality of the impact that social media is having on traditional broadcasting, and it is a fundamental shift in terms of how people consume media.
Take my own case, for instance. I was interested in both the Patriots-Ravens Sunday night football game and the Emmys. They were both running live at the same time. Rather than randomly switch between the two events during commercial breaks, I was using Twitter as my “excitement dashboard.” Whenever something interesting happened on either of the two events–such as a red zone play or an award category that I cared about–I would be socially alerted to it, in real time, and would appropriately switch channels.
Jimmy's prank worked because it had me switch channels to watch the Emmys. Also, his prank triggered thousands of tweets not only with Jimmy's suggested text but also with people's opinions. Some called it brilliant while others found it lame. The key though is that we are now able to quantitatively measure this real-time interactivity. After decades of its existence, we still cannot accurately measure how many people watched the Emmy's on television and what they thought about specific aspects of the show. However, we do have the capability to accurately measure audience engagement on social media. Judging by people's reaction on social media, we now know which of his jokes worked (shooting Don Knotts was not cool), emerging fashion trends based on celebrity outfits (Yellow? really?) and the most memorable acceptance speeches (John Stewart nailed it).
Thanks to big data technologies, we now have the mechanism to capture, measure and analyze the millions of social media activities around major events and gauge audience sentiment in real time. Take for example the continued effort between IBM and the University of Southern California (USC) Annenberg Innovation Lab. Over the last year, we have jointly embarked on multiple “real time” experiments to research how social media and technology can be used by journalists, movie studios and other organizations to understand public opinion and participatory culture. We developed the Oscar Senti-meter tool jointly with L.A. Times to analyze opinions about the Academy Awards race shared via millions of public messages on Twitter. “This project is about identifying 'The People's Oscar,' which means moving beyond pundits' opinion who the winners may be, to understanding who real moviegoers want to see receive the highest accolades of the industry,” said Professor Jonathan Taplin, Director of the USC Annenberg Innovation Lab. “We want to illustrate how the new technologies can capture valuable information and opinions derived from the voices of the influential movie fans.”
The availability of this new technology overcomes traditional challenges associated with processing natural language text at scale. It enables us to take fire-hose style social feeds from Twitter and Facebook and process that in real time so that we can get a sense of the vox populi. It enables us to understand the context associated with each and every tweet and correlate it to a specific entity–such as a movie or an actor. For instance, the Oscar Senti-meter was able to correlate the tweet “Toast her in style with a Norma Jean punch!!” to a positive sentiment associated with Michelle Williams' portrayal of actress Marilyn Monroe in My Week with Marilyn. In order to resolve this tweet, the system had to understand the nuanced usage of language and that Norma Jean was actually Marilyn Monroe's given name! The use of advanced text analytics libraries with entity association rules helps us disambiguate context and understand human sentiment. Also, we need to be able to run this type of analysis at scale. In real-time. During the course of the Academy Awards, over a three-hour window, the Oscar Senti-meter was able to analyze over 2 million tweets to understand what people thought about the programming, advertising and other aspects of the show.
The good news is that this type of technological capability is no longer just a research project. It is now available for enterprises and is part of IBM’s Big Data platform offering. In the following video clip, Professor Jonathan Taplin, describes his experience with IBM’s Big Data platform to power social media based analytics at the USC Annenberg Innovation Lab.
To learn how other organizations are using big data solutions, visit the IBM big data channel on YouTube.