Dawn of the abstraction layer Coming soon to language services
Aggregating multiple offerings
Plane tickets and hotels – more people buy them through aggregators and middlemen than directly from airlines and chains. When you fly to a conference for instance, you likely don’t bother to choose a specific airline such as Lufthansa or KLM. Instead, travel aggregators such as Aviasales, Opodo, or Skyscanner pick the most convenient flight for you in terms of time, price, and route.
In the world of business-to-business commerce, Twilio is a great success story of aggregating multiple offerings. It brought together telephony and sms providers under one API, and it is now a company worth USD 7 billion. It doesn’t even matter to the user which brand or provider they will actually use for the call within the Twilio infrastructure. They trust the aggregator to ensure it’s the most suitable one. Individual phone operators become less relevant – they become an abstraction.
The lesson is that whenever a class of services or technology becomes electronically available via an API, companies can aggregate and resell multiple brands of that offering. The reseller compares offers and helps to determine the best one. Individual brands become less relevant, abstract.
This business model, which is prevalent in the travel and insurance industries, has begun to work its way into the world of translation.
Let’s look at a few cases of aggregating APIs for translation services.
Inten.to and TmxMall connect to multiple MT providers through a single API. This allows the user to switch to the most suitable MT engine for the language combination that they need. CMS middleware products like Xillio, Beebox, iLangl, and Clay Tablet connect translation systems to multiple CMSs so that the buyer can continue using their preferred TMS even if it does not have the required connectors. Companies that work with multiple translation management systems, like Kaleidoscope in Austria, have developed middleware to simplify the integrating of new functions from add-on products.
|Service or tech||Aggregator||Example vendors|
|Machine translation||• Inten.to• TmxMall• Translate4Eu||• Google Translate• Microsoft Translator• GT Com• Amazon Translate• Yandex Translate|
|CMS integration||• Xillio• Beebox• iLangl• Clay Tablet||• Connectors to AEM• Connectors to WordPress• Connectors to Drupal• Connectors to Typo3|
|TMS API||• Proprietary middleware at Kaleidoscope• Proprietary middleware at Gemino||• SDL TMS• SDL WorldServer• SDL GroupShare|
Despite some good examples, translation and interpreting technologists are just beginning to think about the abstraction layer. There are dozens of ideas lying around, awaiting tech-savvy entrepreneurs to try and turn them into products. A cursory look at the field shows plenty of opportunities.
There are about 150 translation APIs in the market – Gengo, One Hour Translation, TextMaster, and Tolingo are just a few examples. Each one provides certain language combinations, but none has the capability of providing all combinations. With some effort, a developer could bring them together into a single API that can check for language combinations and pricing. The new abstraction layer could then provide thousands of language pairs at next to an unlimited scale – and for the best price. The same applies to machine translation post-editing and other similar services.
Of course, there are bottlenecks, such as communication with translators. Buyers who want to maintain a level of control over the profiles of the people who actually do the translation will not be able to use this approach to its full extent. More likely than not however, the next generation of APIs will have to address this challenge.
Remote interpreting is making strides but most of the new companies provide only the technology platform. They struggle to get enough interpreters into the system. In the future, smaller call centers or “home studios” with a few interpreters could become commonplace, and a single API could unite them into one network. Imagine being able to instantly get an interpreter in any language 24/7, 365 days/year from a network of distributed call centers around the world – all of that infrastructure available at the best possible cost to any company that sells to the end buyer. All they would need is a single API to rule them all.
Today, the world’s terminology resides in hundreds of public dictionaries and corporate term bases. Writers and translators need to choose from dozens of fragmented sources in order to find the correct term among multiple available options. The next generation of terminology management will be online, aggregated, segmented, and inside any CAT-tool or authoring environment from Word to Oxygen Editor.
Speech recognition and synthesis technology are so important for search, chatbots, and virtual assistants, yet they are only developed in a few languages in many different silos. Finnish speech tech is developed in Finland, and Latvian comes from – you guessed it – Latvia. Yet Alexa only speaks 21 languages. Other similar apps speak even fewer.
A company that wants to create an automated subtitling or transcription service has to rely on multiple providers. In the future, there will surely be an abstraction layer for speech recognition and an API for hundreds of synthesized voices.
The ultimate form of commoditization
Aggregators want to control the relationship with the customer, making individual providers and their respective technologies less important. Existing providers should see abstraction as a threat to their business model and develop ways to retain customer loyalty in case aggregators really start winning market share.
Technologists with resources and an understanding of customer needs should attempt to exploit the abstraction layer model. Other service industries have been transformed by this development. It is just a matter of time before it comes to language services.
Stay up to date as Nimdzi publishes new insights. We will keep you posted as each new report is published so that you are sure not to miss anything.