Localisation pressures mount: AI and ML begin to come of age

  • Hannah
  • May 20, 2022
  • Comments Off on Localisation pressures mount: AI and ML begin to come of age

Household names such as Netflix have been – if anything – even more exposed to this overall trend, as their initial market success has resulted in stagnation – and for some, reversal – of user growth, resulting in more pressure on developing markets.

Business drivers continue to emphasise localisation

An Ampere report from late 2021 illustrated the challenges faced by the industry at large, using the Netflix lens. The report noted that between Q4 2019 and Q3 2020, North American commissions accounted for 55% of Netflix Original titles, but this share fell to 45% between Q4 2020 and Q3 2021, as international investment by the streamer grew. New markets for the streamer include Poland and Turkey, where localisation is well underway, following a model established in South Korea, Mexico, and South America, where a business model of acquisition of third-party titles, small trials of local commissions and then a wider rollout of original local production has been the norm. This wider rollout was said to begin with a high proportion of movie commissions, then followed by a regular flow of TV-focused activity.

Cyrine_Amor

Cyrine Amor, Content commissioning analyst, Ampere Analysis noted that consumer demand itself is becoming more globalised as a result of this business process, with established English-speaking markets becoming increasingly aware of and interested in subtitled or dubbed global content, adding a further driver to the mix. “So we see that in the US and in the UK, English speaking markets that traditionally are not so accustomed to consuming overseas content there has been over 10 percentage points difference within two years, in [the viewers] watching overseas content. And similarly in a market like France, we see over five percentage points difference within two years and this type of consumption. These changes are mainly driven by SVOD subscribers, so we see that people who subscribe to streaming services are much more likely to change their attitude towards international content consumption.”

AI and ML take some of the strain

These drivers have combined to intensify pressure on content companies to improve localisation techniques and technologies to deliver more volume in a shorter time.

Alphie Larrieu

Alphie Larrieu, Head of Localisation, Astro Malaysia, noted that AI and ML services are offering at least some mitigation against the burgeoning volume. “The advantages of combining machine translation with actual linguists is evident when you start doing volumes. You start to see that you’re slicing your turnaround time, and you’re slicing your time to train and to coordinate with new talent, new linguists and also to pick up a lot of the ‘busy work’ that the market will attract, because of the complexity of changing from one platform to another.

“We do our own house masters as a broadcaster and then we redo a new cut for our OTT platform, then we do a new cut influenced by things like compliance because there are strict laws [in Malaysia]. In this region it’s not uncommon for there to be a whole list of things that we will not show on air, but we will on internet, because there’s a grey area which gives added engagement for our subscribers and our younger audience who want to see the real cut. This all means that we have to do it quickly, the more we can automate and get it through the door [the more likely we are to] beat the pirates.”

Additional metadata a core requirement

Inevitably, when working in a digital environment, it’s not simply a matter of translating the main dialogue, but of capturing as much of the other data points in a section of video as metadata so they can be translated too. Things such as street signs, smartphone messages and book titles all need to be captured if possible, adding exponentially to the challenge – as Mazin Al-Jumaili, ZOO Digital puts it – of ‘glossariasation’. “When we’re getting a full first script, or a first cut of a TV show or movie, we’ll receive as much material from the content provider as possible, that highlights a lot of these sorts of elements, and we can document them and then provide that documentation, that glossarisation. During that initial transcription, you’re actually capturing as much metadata as possible that’s associated with that media, so that the ‘glossariasation’ of anything that’s an on-screen text or graphic can be localised.”

Mazin-Al-Jumaili

While automation technologies in general are improving the turnaround times possible for localising content, many key pain points still stand out – such as the management of versioning in short turnaround projects such as episodic weeklies. Another key pain point is around dubbing, where the requirement to have a global cast of actors available at short notice to dub localised versions is still broadly unavoidable, even with advances in AI tools.

While the localisation market is evolving at a rapid pace, it is clear that there will be much more required of the sector in the coming years, as SVOD players refine their regional offerings and business models. This dynamic sector is certainly set to be a key bellwether throughout 2022 and beyond.

For the full debate and more detailed metrics, themes and talking points, watch the IBC 365 webinar, Localisation for Streamers now.