IRIS.TV Talks Future of Personalization @ LA Machine Learning Group
How much personalization is necessary before we see diminishing returns in engagement and retention?
What is the optimal mix of online and offline operations?
How should technology operations be managed?
These questions and more were explored on September 15, 2014 by two experts from IRIS.TV before a packed audience of data scientists, media executives, entrepreneurs, and press at LA Machine Learning Group’s event at Cross Campus in Santa Monica, CA.
IRIS.TV, the leading in-player video recommendation engine designed to create continuous and personalized streams of video, utilizes a hybrid process to orchestrate computational tasks for learning in order to shrink the latency between the analyzing of history (back end) and presentation of recommendation videos (front end). In order to successfully orchestrate this, the data science and technology engineering teams must work together as one.
In a digital economy moving more towards personalization, consumers expect recommendation services to work seamlessly and intelligently. Publishers and distributors of video are being judged not only on the quality of their content, but also on their user experience. This has resulted in a shift toward personalized recommendation engines to increase engagement and improve retention. Beyond the user interface, the quality of the experience is largely influenced by the ability to present the viewer with the most relevant content in real time.
It is not easy to employ machine learning to optimize user experience. It is a computationally intensive process that often occurs offline. Presenting viewers with adaptive personalized content in real-time requires the use of cutting edge data science and the integration of front and back-end processes. This hybrid approach is neither cheap nor operationally simple.
Dr. Thomas Sullivan possesses extensive knowledge and experience in intellectual property development and is the (co)-author of eight books and multiple patents. Prior to his work with IRIS.TV, Sullivan spent 13 years as a Senior Information Scientist at the RAND Corporation. Joel Spitalnik has a background developing iOS applications for hospital laboratory testing, as well as dynamic websites for arts and entertainment organizations. He is a veteran of the film and television industry, having worked for Disney, DreamWorks SKG, HDNet, and A-Line Pictures.
Together the two explained how data science and machine learning algorithms are used to support recommendations as well as the collaboration with engineering and technology operations.
Sullivan addressed such topics as classification, feature extraction, anomaly detection and how to identify patterns to make inferences about behaviors and preferences of the user. Spitalnik engaged on issues ranging from the curation of video in real time to natural language processing.
“I really enjoyed the presentation. It was a thought-provoking talk about how data science is making the user experience increasingly more personalized and intuitive,” said attendee Shannon Kung.
Attendee Mac Malik added, “It was a fantastic event with great insights. It is clear that IRIS.TV’s usage of big data enables them to deliver a product that not only has great usability but also a great business model to provide accurate video recommendations.”