Home Analysis NENT reveals its use of AI/ML to personalise services – including deep...

NENT reveals its use of AI/ML to personalise services – including deep content analysis

Photo: iStock/aqabizi
Share on

Nordic Entertainment Group (NENT), the content creator, channel owner and service provider, whose brands include Viaplay and Viafree, has revealed its use of artificial intelligence and machine learning to boost content recommendations and personalisation of the UX on its streaming services. Kaj af Kleen, Group CTO and CPO at Nordic Entertainment Group, told Connected TV World Summit that its intelligent ‘machine’ is being used to look deep into content and analyse it for characteristics that can be matched to other content and personal tastes. The ‘machine’ is also used to analyse user behaviour to ensure a deeper understanding of each individual. The magic happens when these datasets are combined.

The deep content analysis is applied to programming that is made or licensed by NENT. A number of techniques are applied: metadata analysis; text recognition; subtitle analysis (across multiple languages); audio recognition, which includes music; face recognition; and other forms of video analysis including scene and activity detection.

Kaj af Kleen gave kisses and cars as examples of activities and objects that can be detected and recorded. Even on their own, subtitles can provide a basis for analysis and comparison, allowing NENT to find similarities in different programmes and so create programme clusters.

The content profile is one half of the story, and as af Kleen explained, his company is using AI to achieve the same deep intelligence about customers. The basics include who is watching and what they have viewed previously.

“Our challenge, like everyone else in this industry, is to provide the user with relevant content at any given time. We use machine learning to anticipate the next move our users will make. We like to think of this as a game of chess, where users make a move and we respond by calculating the probabilities and opportunities to satisfy that user.”

These calculations take context into account – like the time of day and day of the week, and even the weather. NENT is even trying to capture the mood of users, with af Kleen admitting that there are some things you need to ask the viewer.

NENT has been working on explicit post-play content suggestions so someone has a journey straight to another piece of content, but recommendations can be more implicit, like if you re-populate and re-order the content that is shown on a landing page or within a deeper catalogue on a per-user basis. Kaj af Kleen showed examples of this at the London conference.

Smart data analytics means NENT can learn from everything that has happened before, affecting outputs like which content is promoted more proactively to a user, how much of the catalogue is promoted, and how much of a content collection should be shown on the start page, etc. “We are trying all the different places within a service where you can sort, filter and recommend. We try to recommend content all the time. We recommend, learn from it and do it again.”

Relevance, engagement and retention are three important and related metrics that NENT uses to judge its own performance with each user. Engagement includes a measure of completion rates for shows or movies. Retention is judged partly on how often someone comes back for more content.

Share on