Eric Bell, a Natural Language Processing Software Scientist at PNNL, gave a “Pecha Kucha” presentation on the value of analysis across multiple media types. Using this fast-paced presentation style, Eric spoke on exactly 20 slides, each of which was shown for only 20 seconds (6 minutes and 40 seconds in total).
He started by explaining the differences between how we analyze people versus how we analyze companies. With people, we ask questions like “What do they eat? What sort of pictures do they share?” Then we take information derived from different bubbles and mash it together to make a statement. However, we wouldn’t use that method to characterize a company—we see an organization as a much more complex entity with various departments, levels, and sophisticated moving parts.
By way of example, he showed the following Tweet that he recently posted about his cat:
To analyze a Tweet, we typically look at its text, the terms, #hashtags, Twitter handles, and other attributes. We might also try to understand the Tweet’s audience—who is interacting with the post and how that looks over time. We might also analyze the image of his cat via the bit.ly link.
Now we’re starting to learn something about the author and why he made the post. We might also go on to analyze other Tweets and images that this author put up to form a picture of what he likes, what he is interested in, how he interacts.
Eric’s presentation then segued to the question, “Why do we perform analysis in a silo’d, isolated fashion?” We produce reports and charts to make decisions on the basis of data, but we do that by analyzing each individual channel separately and analyzing traditional media apart from social media. He asked whether there is value in trying to bring the diverse data types together. While we suspect there is great value in being able to do this, we see it as a daunting task.
Eric went on to ask whether we can be more broad about the way we think about digital identity and social analysis. Can we blend traditional and social media together? Instead of using an algorithm on a single platform, could we use it on multiple platforms? If we could do this, our decisions would be much more informed, with a deeper and richer understanding of context. In some cases, it might lead us to ask a new question about traditional media or get an insight on something we previously were not able to.
Eric ended his presentation with a challenge to the audience, the “brains in the room.” Can we change the way we have traditionally done social analysis and bring value and impact to the world of social data? If the brains at Big Boulder can’t figure this out, perhaps no one can.