Automation puts forgotten treasures back on our screens

David Muhle, VP sales and channel, EMEA, Ooyala, on the importance of metadata in archives
Publish date:
doctor who

Changing viewer habits and a growing demand for more content has led to increasing pressures on publishers and broadcasters alike. One way to meet this growing demand is to turn to previously created content stored in archives, and plan for how that content can be used in future.

An example of this was seen in November, when the BBC announced a new paid-for digital service that allows viewers access to all their favourite shows from the corporation’s entire archive. The decision was made by the BBC’s director general after speculation that the BBC may be facing a demographic “time bomb”. In an age where young people are deserting traditional linear television for on-demand digital services, organisations like the BBC must stay ahead of the curve by allowing the general public access to their back catalogue.

However, the precise mechanism for storing, organising and finding terabytes upon terabytes of content is difficult. The market has changed very quickly and streaming services like Netflix and Amazon have shown that millions of people prefer instant, on-demand access to their favourite shows. The streaming platforms have also driven demand for automated content recommendations - based on anything from a viewer’s favourite actors, themes or soundtracks.

While the benefits to publishers are clear, and getting the right content to the right audience will always be important, it does create a headache for those challenged with organising the content. While best practice can be put in place moving forwards, publishers have currently not even digitised their archives and, even for those that have, may only have added a few basic pieces of data, such as air date, series name, episode title and (in rare instances) a synopsis. Most archive facilities, however, do not have this.

The problem with no metadata
Archive facilities that concentrate on the finished output and constituent assets tend to have no metadata at all. Archives of clips (unfinished content) from events, for example, may not even have data beyond some camera related metadata. Even for sporting events that may have taken place in the last five years, there is often no way to access these short clips in order to put together, for instance, a compilation of Olympic winners.

For those that do have metadata, there is the issue of where that resides. If this is collated in a database or excel sheet, it is unlikely that the content has click-throughs to the clips themselves, making things almost as complicated as if there were no metadata at all. As such, there needs to be a way of adding metadata to this content and exposing existing metadata in a single searchable form. Getting useful deep metadata is, however, prohibitive in both time and cost unless you can monetise it easily, and beyond just being a pain it can have serious implications on real business decisions.

Tips for best practice
The solution here is AI or machine learning. By utilising the two, a streaming service can create an audio transcript, face detection and identification of actors in that programme, a recognition of who spoke when, recognition of text on the screen, scene detection, sentiment and much more, to search through its archives seamlessly and quickly. All this metadata is then available to be used by the customer. It is then possible to use this metadata to create recommendations based on who is in the content and the style of the content, amongst other things.

It is an extremely powerful tool and, although it is not yet without fault (for foreign language programmes it is only 90 per cent accurate), the manual overheads to an organisation are significantly lower as a result, and refine the job of the metadata worker. Of course, this stat will inevitably increase over time as the machine begins to better understand your content.

This means an archive suddenly has real value to its owner, allowing them to find the content they need quickly and distribute it accordingly. Perhaps it is the Queen’s jubilee and you need to quickly put together a compilation of clips of her used in the past – now you can seamlessly. The ability to find archive content from fifty years ago as easily as you could from three years ago is very powerful and exciting for old-school broadcasters.