The Blog

 

When fans watch a sports match today, they toggle between social media and other messaging apps. This distraction has led to fans watching a greater number of games, but less of each game.

At Trenity, the objective was to achieve an MVP that would allow these sports aficionados to enjoy a game while also engaging with social and digital content. We built a B2B product – an audience engagement tool called Centify – to enable digital media platforms to integrate social reactions with video or textual content. Our clients would be able to cash in on every minute of content consumed and achieve granular brand targeting in a native format contextual to the content, with a bottomless inventory. As a result, users will spend more time watching a match.

Bridging the sports consumption experience

Integrating a social feed into a sports viewing platform has been attempted a number of times in the past, but with little success. The current frameworks fail to add context to social reactions, which means if a user is scrolling through a feed and sees two consecutive reactions (for example, smile emojis), there is no distinction between what these reactions were based off of. Such a curation of feed with little context adds no value to a user who wants to know what his or her social network is feeling towards a certain sports moment.

Trenity provides two layers of real-time social content:

  1. Macro trends where users can find out which players are being talked about in real-time
  2. The ability to read the reactions within a social network towards the player or game

figure_1

Using Watson to process social content

Watson Natural Language Understanding was the key component of technology we used for our product to process social data around real-time events. Our process was as follows:

  1. Identify tweets around a specific match.
  2. Use Watson Natural Language Understanding to analyze the emotion of the tweets.
  3. Use Watson Natural Language Understanding to analyze the sentiment and targeted sentiment (towards entities) of the tweets.
  4. Associate the appropriate emoji based off of Watson Natural Language Understanding’s analysis of the tweets.

figure_2

With this part of our requirements engineering completed, we began building our MVP. We used Tweepy for streaming tweets and Kafka for event sourcing and event streaming. Watson Natural Language Understanding APIs allowed us to quickly process the data we gathered and collect insights that had context and would be beneficial to sports viewers.

A look at our product

Our product displays emojis about trending players during the match or at an instance of the match (if you are simulating a match from the past) that corresponds with viewer reactions on social media. For example, the following emojis are the emotions displayed from the tweets on a football match. To identify tweets pertaining to a match, we used standardized hashtags, like #CROFRA for the Croatia versus France football game, as well as other permutations and combinations of hashtags.

figure_3

What’s next for us

With Watson Natural Language Understanding, we were able to build a compelling product that gained Trenity admission into Y Combinator’s Startup School. There, we aim to further develop the product with even more features, and scale our business.

The Y Combinator Startup School selected all applicants this year, but assigned mentors to only a few startups – Trenity being one of them.

figure_4