There is no shortage of content available at the fi ngertips of television viewers today. This is not least because, in recent decades, television has exploded from just a handful of channels to hundreds. The only downside to this is deciding what to watch. Indeed, viewers are spoilt for choice when it comes to entertainment content. In addition to the hundreds of channels available through traditional platforms, there are apps from networks and sports leagues, as well as over-the-top (OTT) services, such as Netflix, iflix and HOOQ. This has left viewers feeling inundated with an everexpanding ocean of entertainment. The obvious solution is to improve content discovery for consumers, making it easier for them to connect to the programmes and films they want to watch and on whatever device is most convenient at that moment. However, that’s easier said than done. One big obstacle is that most search solutions today still return incomplete with partial results. It is still too difficult for viewers to find the shows they want by searching for logical keywords, such as the name of a cast member or character. In addition, they have been unable to key in or speak the name of their favourite star and then be directed to the latest TV episodes or films featuring that actor. The answer is for search-and discovery tools to evolve from static, one-size-fi ts-all systems to intuitive, recommendation-based offerings.
Artificial intelligence and machine learning are making waves in the media industry
Advances in artificial intelligence (AI) and machine learning are dramatically improving the underlying search algorithms and metadata. This means more accurate and relevant search results, personalised recommendations and more targeted results for viewers. Television search engines can now track and categorise viewing patterns through intelligence on viewing trends, which enables them to deliver content that’s more relevant and interesting to viewers. This is not only beneficial for viewers, but also for studios, broadcasters and networks that want to fully monetise their content catalogues and maximise their revenue. These players are heavily motivated to bring the most compelling content to viewers. However, until now, these large untapped back catalogues have been buried by the abundance of content that the industry has tapped into, and the significant proliferation of distribution channels. The key to more intuitive search and better discovery is deep, real-time and regionalised metadata. By using enhanced metadata powered by AI and machine learning, entertainment providers can make their entire catalogues more searchable and discoverable. They will continually assess the popularity and relevance of entertainment in the real world so that discovery systems can anticipate consumers’ interests and merchandise back catalogues more effectively by referencing related content. As such, entertainment providers will be able to increase viewership by presenting more relevant content to their audiences more regularly.
Metadata as the foundation of television personalisation
The significance of enriched metadata and linking is that it can help surface the most relevant content in real time and at the right moment. It provides discovery systems with a deeper knowledge of entertainment content by identifying relationships between content and keywords – such as ‘Premier League,’ ‘Manchester United’ and ‘Romelu Lukaku’ – and, most importantly, understanding the strength or weighting of those connections. Trending data and algorithms enable more contextually relevant discovery by assessing what is happening in the world at any moment and relating that to entertainment content to anticipate what viewers may want to watch next.
For instance, the ‘Harry Potter and the Cursed Child’ stage play recently won nine Olivier Awards. Since J.K. Rowling was mentioned in trending news stories around the stage play and because of her name being heightened in the press, users reading about the stage play were then linked to all the other Harry Potter fi lms, soundtracks, books, actors, etc. Importantly, they were linked through to ‘Fantastic Beasts and Where to Find Them,’ which was being released in various formats in the UK at the same time. So, a feed of trending entities and topics allows a continuously changing and relevant set of content to be discovered. Enriched metadata and weighted keywords combined with entity linking are key to delivering relevant search results. With AI and machine learning, real-time trending data is used to surface entertainment content based on social media and current events. This enables studios, networks and broadcasters to monetise their catalogues more effectively, because their discovery engines know when specifi c fi lms, television shows or celebrities are trending and ensures that the information related to that content is current. For example, during the Olympics, broadcasters could provide real-time updates on personal bests and the number of medals won for each country.
Machine learning powers conversational search functionality AI and machine learning technologies are now driving the media industry away from paper TV guides for discovering content and towards conversational search and voice features. Speech will play a key role in advanced interaction, as the television experience grows more visceral due to the development of Internet of Things (IoT) technology in the home, such as the growth of speech-centric products like Google Home, Alexa and Siri, for example. As such, viewers will increasingly expect highly personalised, voiceenabled conversational interfaces with sophisticated learning engines and robust user profiles that can anticipate intent and connect viewers with the desired content using natural language.
For example, when viewers are searching for a sporting event, such as the Wimbledon tennis tournament, they should be able to search for matches simply by stating their preference. “Find the Federer match,” for instance. Even if the words “Wimbledon” and “tennis” are never mentioned, the system will know that Roger Federer is a tennis champion and that Wimbledon is a high-profile sporting event at that moment and, thus, the optimal search will result.
When using a conversational search application, users can ask naturally spoken questions and follow-up queries that the technology will understand. This way a user can engage in a normal, free-flowing dialogue, with the voice system responding much the same as an intelligent person in a conversation. To be fully functional, voice technologies must be backed by sophisticated search capabilities, such as dynamic, semantically-linked knowledge graphs coupled with deep metadata. By building such voice and search technologies effectively, consumers can expect to reap the rewards of fast, accurate and intuitive voice content search, and the industry can expect the TV guide to be much more interactive and hands-on.
A more intuitive and convenient user experience
Enriched metadata also leads to greater personalisation and convenience. For example, saying “Reese Witherspoon” and “lawyer” into the discovery engine, but without naming the actual film, will still generate the answer “Legally Blonde.” When the discovery engine doesn’t make the link, it’s often due to the data providers having limited metadata and image assets. This is often the case with the sports discovery experience, where the end user experience is often disappointing – it’s not as rich and satisfying as film and TV discovery experiences. With machine learning and AI solutions, such technologies will be able to provide real-time updates, including sports highlights by game, league, player and season, as well as offer related entertainment content across various media types.
The goal is to realise the benefits of moving from manual curation and QA, which is the current industry standard, to a hybrid model that combines the best of editorial expertise with the scalability, accuracy and dynamism of algorithmically-supported machine learning. For example, humans tend to be able to describe a TV show or a film with emotional descriptors like its mood far more easily and accurately than machines. When that editoriallydriven descriptive information is added to knowledge graph-based datasets, it enhances the system’s understanding of entertainment content and its ability to identify meaningful relationships to fuel more relevant recommendations.
AI and machine learning will continue driving the media industry
Ultimately, the media industry is only at the start of the AI and machine learning revolution that’s changing the way consumers search and discover content. Advancements in AI and machine learning technologies means studios, broadcasters and networks can now use enhanced metadata from their large, untapped back catalogues to consolidate, normalise and structure that data to create better entertainment experiences for viewers. By providing better search, recommendation and voice-enabled discovery features, consumers will be able to easily fi nd their favourite TV shows, fi lms, music, actors and games – as well as discover new ones – while studios, broadcasters and networks can deliver a unique experience, increase consumption, build loyalty and, ultimately, drive new revenue.
