Aditya Chaganti

Aditya Chaganti

Machine Learning Engineer

The Problem

Automated curation of User Generated Content (UGC) is one of the hardest problems that our customers have had to tackle. UGC media is eclectic – it could be media submitted from a football stadium, or one submitted by a person in the middle of the woods. Producers and content moderators face the arduous task of sifting through media from disparate sources such as these, and picking out the best among them for their broadcast, social media, and digital feeds. As the volumes of UGC media submissions increase, this can very quickly become a needle in a haystack problem.

The Solution – Burst Search

Artificial Intelligence (AI) assisted automatic curation of UGC media allows producers to quickly find the most compelling content submitted by their users, and monetize it with ease. Here at Burst, we have been developing AI driven approaches to solving this problem; models that quickly scan media sent in by users, and glean information from it.

Automatic curation, therefore, saves time and energy while allowing customers to maximize the engagement value of the media that is shared across platforms. Achieving this optimum remains the core focus of the AI team here at Burst. Burst Search is a scalable solution that enables automatic curation, the development of which was a big step toward enabling users to leverage these capabilities.

What can users do with Burst Search?

Search is an integrated solution that combines intelligent curation with an easy user experience, allowing for minimal time between when the best content is acquired, to when this content is ready to be pushed out to social media, widgets, and broadcast streams using Burst’s many integration capabilities. Users get quick access to information extracted from newly uploaded media by Burst’s Artificial Intelligence models (object detection and audio classification for example), as well as information obtained about more mundane aspects such as the media type, time of upload, and the location at which the media was captured. Upon upload, this information is extracted from the media and made searchable in near real time.  Users can also perform similar searches on “bubbles”, which on the Burst platform, refers to a collection of media that are all related to a common topic of interest. Examples of bubbles could be a bubble for breaking news, an impending storm, a back to school bubble, or a Halloween bubble, etc.

Use Case – National Pet Day

National Pet Day is approaching, and the local news station would like to run a promotion with a social media campaign, and a TV segment covering local shelters and their efforts to find the rescues in their care a new home. The segment will also air several UGC media submitted by the station’s viewers talking about the joys of adopting a rescue and having a pet. After putting out the call to action (A social media post or a broadcast prompt asking viewers to send in their best videos), focus shifts to other work. The calls to action prompt a large number of video submissions from viewers, but also result in some fans sending in media that aren’t quite related to pets (media that probably shouldn’t make it to broadcast TV). Right before the show airs, under normal circumstances, the content moderator is faced with the task of sifting through the hundreds of media and finding the best among them. Pressed for time, this could also become a matter of picking the top 10 media that are acceptable. That could mean that some quality content like this video of a goose that one of our customers received from a user never sees the light of day. This video was viewed 19 million times, shared 320,000 times, and drew 15,000 comments from people around the world on Facebook.

A situation like the one above is one that is all too common in TV stations around the country. Having access to AI enabled search capabilities would allow moderators and producers to save huge amounts of time in curating media, and picking the best among them. In the use case above, a search for “pet”, would bring up a list of the best UGC media in the bubble that contain pets in them. Further, the best among these pet media appear at the top of the search results, whereas the less engaging media appear at the bottom. At that point, a few clicks are all that is needed to push the best automatically curated media to digital, or broadcast workflows using Burst’s integrations.

Curation At Your Doorstep

We understand that different users have different areas of focus. While the use case discussed so far focuses on TV stations, we work actively with users across multiple domains to ensure that our capabilities best serve their specific needs. The same philosophy applies to search as well. It is an evolving product that we are excited to be working on, and improving every day based on feedback we get from our active users.  Search allows Burst to deliver automated curation to the doorstep of customers that not only save time and resources from using it, but also benefit from our state of the art domain expertise in Artificial Intelligence.