Digital Marketing: HP Autonomy at the Sundance Film Festival 2014 - Part 5

Throughout my blog series about our Social Media & News Command Center at Sundance, I've talked about where to listen, what to listen for, and how I physically chose to present the content. I also mentioned that HP ExploreCloud installations fortunately don't require one thing that all the others do: a team of analysts—between you and the big screens—reading and making sense of the data. ExploreCloud replaces those analysts with HP Autonomy's human information engine, HP IDOL. IDOL truly is amazing, and I would encourage anyone to read more about it; but today, I'll just go through some of the highlights of how ExploreCloud uses that engine.

 

 

Meta-Data Extraction and Normalization

It all starts as soon as I click save on that first listening rule. As the social and news content flows in, the system starts extracting numerous and varied pieces of metadata: number of Twitter followers, Facebook likes, YouTube tags, Tumblr re-blogs. These hundreds of pieces of semi-structured information are normalized so that we can look at all of the data at once, not in silos where you could miss the story.

 

Education

Next, IDOL digests the content—a 140 character Tweet or a 3 page news article—and educes out what is being discussed. It figures out, in context, things like what people are being talked about, what places are mentioned (regions, cities, countries, etc.), and what companies or organizations are referenced.

It determines the authors' language and dialect (it’s just as important to distinguish between US, British, and Australian English as it is Mexican and Castilian Spanish). Then, it identifies positive and negative "vibes" from the individual snippets of text that are used favorably or unfavorably. Again, context and regional language matters. "Can't wait" usually has a negative vibe or connotation in an order-status email, but is nearly always positive in a Tweet among friends.

 

Context

Subsets of the content are then automatically sorted into different “Projects” which put the conversations into context. ExploreCloud customers use Projects in many different ways. Some use Projects to focus on individual product lines, separate support team tasks from sales team initiatives, or segregate secure content by authorization level.

 

With some installations, Projects are used to filter out “distracting” content like profanity, heckling and questionable images from the public display, but still keep all the content for review.  For example, at Sundance we’re using Projects to separate out what people are saying about Robert Redford–the man (and they say a lot)–unrelated to the Sundance, so we can focus what people are saying in reference to the festival and its films.

 

Conceptualization

One of the most valuable IDOL functions comes next. It identifies and prioritizes “concepts” within the individual posts, the important snippets of text that are being talked about. Then it finds relationships within the data, surfaces trends and issues that develop, and pushes them to us. HP ExploreCloud lets you interact with those concepts in a variety of ways. Here are you can see word cloud, node graph, and map versions of Concept widgets (what you can’t see is that ExploreCloud lets you interact with them):

Sundance2.png

Sundance3.png

 

IDOL is not just counting words; it’s finding the relevant Concepts and then presenting them based on importance (using a combination of Bayesian Inference, Shannon's information theory, and pattern matching). By understanding information the same way humans do - identifying and prioritizing concepts in context – ExploreCloud replaces that row of analysts and avoids reliance on manual tagging, keywords, or metadata.

 

This entire process I’ve talked about here happens in seconds; so someone standing in front of the wall can take a selfie, tweet it with a few ironic words, see the picture scroll by, and also see their text influence the conversation instantaneously.

 

The data tells the story.

 

#HPDMB 

Leave a Comment

We encourage you to share your comments on this post. Comments are moderated and will be reviewed
and posted as promptly as possible during regular business hours

To ensure your comment is published, be sure to follow the Community Guidelines.

Be sure to enter a unique name. You can't reuse a name that's already in use.
Be sure to enter a unique email address. You can't reuse an email address that's already in use.
Type the characters you see in the picture above.Type the words you hear.
Search
About the Author
Martin Cuéllar, Director of Product Marketing for HP Software, is a Subject Matter Expert within HP Autonomy Emerging Technologies. Martin’s...


Follow Us
The opinions expressed above are the personal opinions of the authors, not of HP. By using this site, you accept the Terms of Use and Rules of Participation