Putting the puzzle together
For several years, IBM has used radio frequency identification (RFID) tags embedded in guests’ badges to track which sessions they attend at a conference. After the event, this data has been used to analyze the popularity of each session, and make more effective plans for the next conference.
Dupard comments: “The company that provides the RFID technology, AllianceTech, also has a big part to play in this story. At KollaCode, we have been working with them for some time to help them pioneer new analytics on the IBM Bluemix platform. And it was their CEO, Art Borrego, who introduced us to Francis in the first place—so the serendipitous meeting that inspired this whole project would never have occurred without AllianceTech.”
Darice Mathis, Co-Founder of KollaCode, explains: “We realized that there was a lot of potential to use the AllianceTech RFID data to track how guests are moving through a conference in real time. If we can see that they’re in a certain room at a certain time, then we can check the agenda and find out which session they are attending. We can then send them messages—at the moment, via Twitter direct messages, but in the future, perhaps via the conference app itself—to give them some recommendations for where they might want to go next.”
Francis Friedlander adds: “That was the starting point, and then we took the idea further: we wanted to base our recommendations not just on the sessions that people were attending, but the ones they were actually enjoying. So we built in a feature that monitors guests’ Twitter feeds—with their permission of course—and analyzes the photos that they share during the conference.”
The solution harnesses the latest IBM® Watson™ cognitive technology to recognize the content of each photo. It then compares the timestamp of the Twitter post with the RFID data from the guest’s badge and the conference schedule to find out which session the photo was taken in. Finally, it uses a business rules engine to decide what the photo implies about the guest’s feelings about the session.
Dupard comments: “When someone really likes a session, they might take a photo of the presenter standing at the podium, or even focus in on a slide from the presentation that they think is particularly insightful. When they’re bored, on the other hand, they take selfies!
“We have trained IBM Watson to distinguish between these different types of photos, and we can feed that insight into our recommendation engine to guide guests towards more sessions that they’ll really enjoy.”
From a technical perspective, the solution—known as Attendee Real-Time Insight, or ARTIST—is built on the IBM Bluemix® cloud application development platform, which provides simple access to a vast range of services via easily integrated APIs.
In this case, the solution uses a Twitter API to collect tweets and images from conference guests; an API from AllianceTech, the company that provides the RFID systems for IBM’s conferences; another API from the IBM conference team to provide room and scheduling data; and the IBM Watson Visual Recognition API to recognize the content of the photos.
Mathis says: “Bluemix makes it so easy to pull all these services together and co-ordinate them into a solution that really works. We were able to move from the initial idea to a first working version in just three months—just in time to test it out at IBM World of Watson next week.”
At the heart of ARTIST is IBM Cloudant®, a NoSQL database that acts as a cloud-based repository for all the data that KollaCode needs to ingest. Because Cloudant is a schemaless document repository, it is able to handle all of these different types of data and store them in JSON format without the need to build complex transformation and normalization processes up front.
For data that ARTIST needs to analyze in more detail—for example, when it calculates how many people are currently in each room of the conference center by aggregating RFID data from more than 20,000 attendees’ conference badges—ARTIST uses another database, IBM dashDB™. Designed as a cloud data warehousing platform, dashDB provides built-in analytics tools that make it easy to perform complex calculations on large volumes of structured data.
The results of the dashDB calculations can then be correlated with the data held in Cloudant—for example, the categorization of the Twitter images provided by Watson Visual Recognition, and the conference schedule—to give organizers a richer view both of the conference overall.
At the same time, the various events relating to each attendee (such as their current geolocation based on their RFID badge or the tweets that they are sending), are sent to the IBM Operational Decision Manager Advanced Decision Server Insights platform.
The platform holds a “context” for each attendee (a collection of data about the attendee, including their current location and past activities), which provides insight into their individual journey through the conference. The platform uses business rules to detect opportunities for each context in real time, and decides on the best action to take—for example, sending a recommendation for an interesting session that the attendee might want to attend next.
Dupard comments: “The ability to choose different data stores and services for different purposes and link them all together so quickly and easily in the cloud, is a complete game-changer. For the first time, it is possible to develop very sophisticated, multi-component applications without spending weeks setting up development environments, configuring database schemas and developing integration layers.
“As a result, it’s possible for smaller development teams to work on much larger and more ambitious projects—and for larger companies to be more experimental without worrying about risking large sums of money on projects that may not pay off.”
Ahead of its first real-world trial at IBM World of Watson 2016, the KollaCode team is confident that the solution will signal the start of a revolution in the way that large conferences are organized.
Mathis says: “We’re looking forward to seeing what we learn at World of Watson, and we’ll keep refining the solution and adding features so that we have an even more mature product ready for IBM InterConnect next March. We’re sure that people who take advantage of our solution are going to learn more, see more, and have more fun at these conferences than they’ve ever had before.”
Friedlander adds: “Looking at the bigger picture, the solution won’t just benefit attendees, it will benefit the organizers too. Running a successful conference means making the most of limited resources—and time is the most important of those resources. You might have a top executive who is available to meet clients for a few hours, before she flies off to her next engagement—and if her session isn’t fully booked, you’re wasting her time. With ARTIST, we can ensure that those rare opportunities get snapped up instead of going to waste.”
As another example, if a particular session turns out to be unexpectedly popular, and is very oversubscribed, organizers can see it instantly and ask the presenter to do a repeat session the next day. ARTIST can then notify attendees of the update to the schedule, and give them a chance to take advantage of a great session that they might have missed first time around.
Dupard concludes: “The technology from IBM and AllianceTech is helping us get so much smarter about how to make conferences more dynamic and responsive to attendees’ needs and desires—we can really understand what sessions have the biggest impact, and predict which activities are likely to appeal to each individual guest.
“But it’s not all about hard science and profiling—we still hold to our original idea that there’s an element of serendipity that even the cleverest algorithms can’t provide. So for every five recommendations we make, there’s going to be one that’s a little random, a little unexpected. That’s the one that we hope will really open people’s eyes and let them see things from a new perspective. And perhaps that’s the one that will make them want to come back again next year.”