The translation industry is growing, but not fast enough! The amount of content published every day vastly exceeds the amount that can be translated by the few hundred thousand translators in today’s workforce. Machine translation seems to be a part of the solution, yet the world is still waiting for this technology to mature. How do we scale up the translation industry’s capacity for the massive potential opportunity?
As part of the TAUS Annual Conference 2015, TAUS invited industry leaders from both the translation buyer and vendor side to talk about what it will take for the industry to innovate. The panel included the following participants:
- Paula Shannon (Lionbridge)
- James Douglas (Microsoft)
- Marco Trombetti (Translated)
- Aiman Copty (Oracle)
The framework for the discussion was laid out by Clayton Christensen’s The Innovator’s Dilemma, in which he delineates two main types of innovation: sustaining and disruptive.
| Sustaining Innovation | Disruptive Innovation |
|---|---|
| Incremental improvements to existing products and services | Sometimes sudden entry to an industry with novel products and services |
| Pursued from within a company or industry | May enter an industry from outside |
| Tends to be sophisticated, expensive, and complicated | Tends to be simpler and may appear flawed or unattractive |
| Targets existing high-paying customer segments | Targets new consumers at the bottom of a market |
| Generally yields higher gross margins | Generally yields lower gross margins |
Paula Shannon argued that although we talk about machine translation, we often do so in the context of sustaining innovation, not disruptive. The main indicator of this is the industry’s tendency to compare machine translation quality to the quality of a human translation.
Ms. Shannon added a third type of innovation to the discussion: devastating. She described devastating innovation, or “big bang disruption,” as the innovator’s disaster. Devastating innovation wipes out an industry overnight with lower costs, better quality, and new business models.
Where will innovation come from in the translation industry? The panel commented that a medium to large LSP is unlikely to have the resources or runway to come up with truly disruptive innovation. Christensen’s innovation model would suggest that translation providers are so preoccupied with serving their high-paying customer segments with sophisticated solutions that they disregard serving new consumers at the bottom of the market with less attractive solutions solutions that yield lower gross margins.
The panel and attendees understood 5 elements as key to innovation:
Speech – Innovation for the translation industry is going to be about communication, and there is no more direct or personal form of communication than human speech. Speech translation technologies are enabling school children to communicate with other classrooms around the world, across language barriers. The challenge moving forward is not the lack of translation, but the lack of devices, bandwidth, and functionality. (Speech translation can be considered a part of the larger trend of multi-modal translation, which includes speech, images, telegraphic text, SMS, and social media posts.)
Data – In the history of machine translation development, the rules-based model was used until critical mass of data was achieved, enabling the statistical based model. It’s likely that another critical mass of data will enable another leap forward, enabling systems to use artificial intelligence, predictive analytics, machine learning, and sentiment analysis to make intelligent decisions.
Business flexibility – New business models can enable disruptive innovation. The panel considered several ideas: setting up competing groups within a company, being open to collaborate with other companies, putting small stake investments in disruptors (possibly through an industry hedge fund), and supporting the fast horses, meaning being willing to use new technology produced by industry leaders. Why should we tell our customers that we can’t use the latest technology?
Technology infrastructure – The obstacles the industry faces are not translation problems; they are technical infrastructure problems. How will the industry adapt to speech as a source format for translation? How will the industry leverage the coming flood of data? How will the industry adapt its client and vendor-facing interfaces? As an example, the infrastructure of the internet was a great accomplishment. But internet infrastructure existed for many years before the jump to mainstream consumer use, which was mainly a success of interface.
Technology talent – The biggest contributor to successful innovation is assembling the right people with the right skills. Tech companies need the best engineers they can get. This means that even for localization engineering positions, the best engineers are scarcely ever found from within the localization industry. The top skills needed today are in computational linguistics, NLP, data science, and SEO.
TAUS Innovation Excellence Contest
TAUS invited innovators from within the industry and from without to compete in the TAUS Insider Innovation Contest and the TAUS Invader Innovation Contest. The winners are profiled below.
TAUS Insider Innovation Contest Winner – MateCat
MateCat is a free, cloud-based, open-source translation tool for LSPs and translators. Instead of saying no to their customers, users can outsource translations with a single click. MateCat features the functionalities of standard CAT tools plus three main advantages: cloud-based translation memory with over 15 billion words and machine translation, one-click outsourcing, and business intelligence from statistical data collected from translations.
By allowing other companies to build on top of the open-source software, MateCat is bringing innovation to the industry in the form of openness and transparency.
TAUS Invader Innovation Contest Winner – Unbabel
Unbabel is TaaS, or Translation as a Service. In Unbabel’s workflow, everything starts with a pass through machine translation. Then, AI assists by deciding which of Unbabel’s 34,000 translators gets to edit what. Once the task is assigned, translators can complete their tasks on their mobile device not by entering in text, but by interacting with a translation graph inside Unbabel’s app. As the translation is being carried out, Smartcheck AI uses real-time quality assessment to help translators do a better job.
Unbabel is a translation subscription service which covers any type of content in 50+ language pairs. Unbabel’s TaaS can be integrated via API to a range of services.
Unbabel is bringing innovation to the industry with its unique MT/AI/crowd-powered workflow, its subscription-based payment model, and its API integrations.
5 minute read