In this fragmented world of over the top (OTT) video services, viewers are overwhelmed with the endless amount of offerings that often serve up less relevant, uninteresting and impersonal content. Current recommendation systems are not up to the expectation of the user and ads tend to be less relevant or not personalized.
With radically changing consumer and subscriber expectations, it’s crucial to understand viewers’ interests and offer the right content at right time with highly personalized ads and suggested content. The traditional logic of using viewing history, device details, location and demographics is inadequate. Also, the search triggered by the viewer is not effective if, for example, a search for latest thriller movies lists old titles or movies from other categories.
The overlap between video content and viewer obligation is less varied. This means the metadata of a video file is not enriched enough to meet the growing demands of the viewers.
If content is the king, metadata is the gold.
AI to the rescue:
The metadata enrichment process should be intelligent enough to handle all types of requests from the viewer. For example, instead of voicing out “thriller movies,” the viewer should be able to search something like “romantic thriller enacted by Suriya Mohan Sekaran and by mentioning <<specific scenes>>.” To achieve this feat in media content search, the process of metadata enrichment must be more dynamic. Specific details like social factors, celebrity identification, keyword extraction, popularity rating, sentiment analysis and emotional detection must be carried out at the time of post-production using the power of AI. Combined with human curation, this AI-powered metadata can create more satisfying video discovery and highly personalized user experience with the right recommendations resulting in increased user engagement and customer satisfaction that in-turn helps reduce subscriber churn.
Augmented metadata is an efficient method to enrich existing or create new metadata by analyzing the video scenes using image recognition and the closed-captions available within the content.
Image recognition uses AI to identify objects, people, places and actions in images or scenes. Augmented metadata is reinforced by machine learning and paves the way for disruptive innovation in the field of content discovery.
An AI-powered content metadata system is very important for the OTT service provider to engage the viewer with more advanced search or granular content discovery experience. Metadata extracted can be used to build powerful engagement experiences with highly personalized recommendations and help improve the value of videos by delivering more relevant ads using the extracted insights as additional signals to the ad server.
Contextual metadata deep video analysis helps to identify main refrains within the content, the leading themes and even the personality qualities or behaviors of different characters based on language analysis. Frame-by-frame analysis helps detect what characters do and portray through the feelings. With this level of metadata enrichment, the viewer will be able to search a movie with a specific scene. Based on precise content tagging, if any channel is broadcasting the interested scenes the viewer requested earlier, the OTT app can display the thumbnail of the channel, creating an interest for the viewer to watch more relevant content.
Machine learning, a subset of AI, enables content to be enriched with additional data descriptors such keywords, popularity ratings and emotion detection. Machine learning can also recognize the trending video dynamically and help ensure the content is served to the subscribers thereby monetizing the content in a timely manner.
Deep learning is a subset of machine learning composed of algorithms that allows software to train itself to perform tasks like image recognition by exposing multilayered neural networks to existing data. Image segmentation is an important method in deep learning that help understand the context where various objects in a given scene are, and to understand their relations to other objects.
With the power of AI, video can be analyzed scene by scene, and the audio of different languages, sub title and closed captions can be interpreted with useful insights for enhanced viewer experience.
More uses of AI and machine learning for enhancing OTT content:
- Videos can be categorized based on the micro genres and the language that enables fine-grain search
- Auto subtitle creation and translation: Leveraging AI capabilities to create subtitle or closed captions (WebVTT, TTML, or SRT) automatically by decoding the texts in video, detect spoken languages and support multi-lingual content. Dynamic translation of existing subtitle to other languages (translate any text, including spoken text, written text, and keyword).
- Texts (closed-captions or subtitles) can be translated to different audio languages and can be made available to the viewers with visual impairments. It also provides a transcript of conversations and descriptions of other noise for the benefit of viewers who have hearing impediments.
- Indexing spoken words and faces can enable the search experience of finding moments in a video where a person spoke certain words or when two people were seen together
- Scene level contextual metadata provides the opportunity for Contextual Advertising or Contextual Commerce (for example, it can display the products used by the celebrity and take the viewer to e-commerce site). The more accurate and rich description of the character name, player or artist name, keywords, topics at scene level provides better control to the advertisers to pick the right slots that best fit their target audience and better reach of brand experience.
- Scene analysis based on deep search capabilities and with the combination of facial recognition, viewer can search for an episode with character name and specific scene
- Second Screen Application: Enhancing the metadata at frame level shall help facilitate interactive session with viewers (for example, the user can be provided with off screen details to enable them viewer to plan for a holiday trip to a precise location in the movie scene with special offers).
- Personalized Ad Insertion: Exclusive contextualized information and advanced analytics help to insert personalized ads that increase the relevance for the advertiser’s brands and products
- Auto Dubbing: Leverage AI capabilities for audio dubbing automatically by creating audio files dynamically
Why AI for OTT monetization?
In order to serve the right set of audiences with relevant content, it is important to enrich the metadata of all streaming content with more information beyond today’s standards. Deep learning helps to analyze video to a level where more useful insights can be extracted in-depth and leveraged efficiently for OTT monetization. This approach of deep video analysis, audio, text interpretation using AI coupled with predictive analytics enables faster content discovery, provides more accurate information to the right audience at the right time, and creates an altogether more enjoyable user experience.
Altran is the premier solution-oriented technology service provider to the media industry. Our long history in multimedia, our prime focus on our customer’s needs and unchallenged legacy of innovative approach enables us to deliver best-in-class solutions for OTT service providers, broadcasters and media companies.
To find out more about how Artificial intelligence can power your video monetization contact an Altran expert today.