An interesting new social TV startup has hit the market. Jinni “is a mood-based discovery engine that is accomplishing for TV and film what Pandora achieved for music – providing content recommendations based on tastes and interests and incorporating social features.” The platform has already partnered with the largest telecommunications company in Belgium and is looking to expand in the U.S.
One of the greatest questions that has been asked in the living (besides where has that lost remote gone) has been, “what are you in the mood to watch?” Jinni aims to build a social TV discovery engine that’s devoted to tagging entertainment content so you can type in a word like “depression” and get a slew of content with depression themes. Accurately, when we searched this term for real (and filtered by TV), Six Feet Under, In Treatment and The Sopranos all showed up in beautiful big tiles. When you click on one of the results and select “watch it”, you’re given options to choose to stream it from Amazon, Netflix and more. We interviewe CEO Yosi Glick about the platform.
Lost Remote: How did you come up with Jinni?
Yosi Glick: I spent over a decade implementing TV middleware for cable companies. Time and again teams would set out to build a better guide only to end up with the same old grid that has been around for decades. I realized that the reason attempts at improving the content discovery experience were failing were because the guide can only be as good as the underlying data.
The same generic metadata (title, channel, airing time, genre, etc.) was being used to build all the guides. This limitation is logical because metadata was designed for content cataloging, not discovery. Metadata is content-centric, serving the purpose of content storage. My dream is a user-centric guide that serves the user by enabling a natural, enjoyable analog content discovery experience.
The Jinni breakthrough began with the understanding that entertainment is special. The same old search and discovery tools that were developed for browsing the internet and cataloging content simply can’t capture the rich and emotional experience of entertainment. Our challenge was to create an entirely new classification system capable of reflecting content the way it is created, the way it is experienced by humans. So we assembled a diverse team of film, media and internet professionals, engineers, and world-class scientists in Natural Language Processing, Machine Learning, and Sentiment Analysis. Together we set out to develop a new approach to categorizing and discovering TV shows and movies. The result is a sophisticated patent-pending automated tagging robot and a rich, intuitive set of discovery tools that make finding content as fun as watching it.
LR: How does it work?
Glick: Jinni has developed a proprietary Entertainment Genome that contains over 2000 “genes,” or semantic tags. 30-60 genes are assigned to each title to describe plot, mood, style, setting, soundtrack and more. This means that Jinni’s service is based on natural language descriptions of content.
New titles are automatically indexed by analyzing reviews and synopses, using proprietary Natural Language Processing technology and algorithms.
LR: How is it like Pandora for TV?
Glick: Pandora is a wonderful semantic discovery service that uses a team of human music experts to manually tag content using a set of about 400 total tags. Jinni does the same thing automatically for TV and movies. Our proprietary technology extracts a richer set of tags per content item (~30-60) and the process is fully automated, making it very fast, consistent, accurate and cost effective.
LR: How will it make TV more social?
Glick: TV is an inherently social activity. We watch TV and movies together, discuss what we’ve seen and get recommendations from friends.
- ‘Watch Together’ enables multiple users to find TV shows and movies that suit their combined Entertainment Personalities by suggesting content they will enjoy watching together.
- ‘Neighbors’ generates recommendations based on content viewed and enjoyed by other users with similar Entertainment Personalities. Jinni is able to cut through the ‘noise’ of other social recommendations that simply suggest you watch what your friends are watching, by identifying friends and users that have very similar taste in TV and movies.
LR: What’s your background?
Glick: I have over a decade of experience in TV and entertainment. Prior to founding Jinni, I served as VP Marketing and Business Development at Orca Interactive, and before that was at Oracle.
LR: Anything else?
Glick: Regarding “Semantic-discovery across multiple domains” – Jinni cracked the code of video discovery with the creation of the original Genome. Now we are beginning forays into additional entertainment domains, such as books and games. This is very exciting territory and we are getting outstanding results from pilot research.
Regarding “Semantics-driven Voice-controlled guide” – Sci-fi movies and shows have long promised a future in which people can seamlessly interact with technology, by just speaking naturally. Apple’s launch of Siri adds new excitement about voice-controlled command, but for voice command technology to be natural and accurate the systems will need the ability to derive rich semantic meaning from natural speech. If all they do is let users say ‘channel up, channel down’ instead of using their thumb on the remote this technology will sorely disappoint. But if paired with semantics-driven NLU (natural language understanding), the once sci-fi dream of telling your TV you’re in the mood for something “uplifting and humorous about best friends I can enjoy with my girlfriend” will soon be a reality.
NLU (natural language understanding) will change the future of intuitive interaction with the TV – and Jinni is perfectly positioned to lead this market.