22 - 24 March, 2017 in Amsterdam (NL)
Brainstorming the collaborative commons of the global translation industry 2017 – 2021
Note: Participation is on invitation only
TAUS expects that automation in the translation sector will accelerate in the next five years. We foresee a rapid adoption of automation, both in translation production (machine translation, intelligent TM) and in translation management functions such as project, quality and vendor management (machine learning). This automation will bring along opportunities and challenges. Opportunities are of course the increases in efficiency. Challenges we’ll find in the changes in jobs, sharing of data, getting intelligence from the data, working in the cloud. Opportunities also exist in the innovation made possible through translation automation. This comes to us through convergence: convergence of technologies, such as machine translation and speech technologies, but also convergence of markets and business models: freemium and paid, consumer and business, ‘crowd’ and professionals. Translation is on its way to become a utility. Buyers and providers of professional translation services need to assess their strategies in light of a changing landscape. The TAUS Industry Summit aims at delivering a roadmap for innovation in the translation industry in the next five years.
The objective of the TAUS Industry Summit is to deliver a Translation Industry Innovation Roadmap 2017-2021. The purpose of this report is threefold:
- Influence the ecosystem and stimulate sound innovation across the translation industry
- Provide a reference for strategic business planning in individual companies that buy and provide translation services and technologies
- Identify services and features to be developed and offered by TAUS as part of the membership program
The full report will be available to all TAUS members.
Participation and Representation
The participants in the Summit represent the larger TAUS member organizations and are empowered to take decisions and represent their companies with regard to the themes and the collaborative industry actions discussed at the Summit. In addition to the members of the TAUS Advisory Board, there will be room for a maximum of twenty other participants in the Summit.
Participation in the Summit is by invitation only. Participants are requested to register by sending an email to Jaap or Anne-Maj and pay a fee that will cover the costs for meeting facilities, catering and dinners and some additional organization overhead. Estimated costs per participant: € 1,000. Participants pay for their own travel and hotel expenses.
The briefing below on the six innovation themes is meant to provide context to the discussions at the Industry Summit, not to limit them. The themes may overlap and influence each other.
1. Machine Learning
Machine Learning may be seen as the self-learning of a new generation of MT engines. It is being said for instance that the new Neural MT technology does not require programming or training in reordering sentence structures for syntactically different languages. Neural MT learns to do this autonomously as long as we feed the data. Since we have a separate machine translation theme on our agenda, we would like to focus the machine learning discussion on the automation of management tasks that tend to be reserved for human operators, like resource allocation, content profiling, terminology identification and harvesting, quality prediction and estimation, MT engine selection, translation assets selection for optimized leveraging, translator evaluation, training translators based on error statistics. As illustrated by the broadly shared stories on how the best Go! and poker players in the world are beaten by machines, machines can often indeed do a better job than humans. As long as we feed them the (right) data.
Machine learning, or algorithmic management, or robotization, as it can also be referred to, in the translation sector is developing very quickly. Not only do we expect that existing functions - such as project, quality and resource managers - can become much more productive, we also foresee innovations as a result of the business intelligence that we will be gathering.
This theme links with Machine Translation, Quality Management and Data Cloud.
2. Machine Translation
The recent new advances in Neural MT and Adaptive MT make managers in the translation industry wonder whether a straight-out rejection of MT is sustainable for much longer. Estimates of the percentage of post-editing MT on the total translation production vary wildly, simply because we do not always know whether MT is used or not, authorized or not authorized. Some customers insist that MT is being used, others insist on just the opposite. As a consequence the industry is torn by a mix of ethics, sentiments and business incentives. A deliverable from the Industry Summit could be a ‘legalization’ or fair treatment of MT technology. This would be an important step towards transparency on evaluation, productivity and pricing, and an important recognition and impulse for the relative small group of independent MT development companies that the translation industry depends on.
This theme links to Data Cloud, Academy, Machine Learning and Quality Management.
3. Quality Management
Quality used to be a static concept in the translation industry. Undisputedly translation buyers would expect human, i.e. good quality translations. Now, in many cases, users seem to be happy to accept good enough machine translations, especially when they are delivered in real time. The TAUS Quality Dashboard with the underlying MQM/DQF metrics is getting traction as a standard approach to quality evaluation. It can help to differentiate, measure and benchmark translation quality according to agreed content profiles. It potentially takes a lot of the subjectivity out of the quality management function. Linked to the Machine Learning and Machine Translation innovation themes, DQF helps to aggregate data that support decisions on the training and selection of MT engines and the matching of resources and translation assets. A deliverable from the Industry Summit could be a mapping of the business intelligence that stakeholders are most interested in obtaining from DQF and a roadmap of new features. This theme links to Machine Learning and Machine Translation.
4. Data Cloud
From the ideas submitted in the consultation among the TAUS members, it seems that it is time for a new generation Data Cloud. Suggestions were made to create a dynamic data marketplace where companies can bid, and perhaps put up a joint bid, for translation data in lesser resourced languages and domains. The sellers of such data could receive a commission for producing the data or making their data available. This could be the answer to a growing need for data in for instance African and Indian languages and in industry domains that are currently not well covered by the TAUS Data Cloud. Another idea is to establish an industry-shared data pooling platform that seamlessly integrates with the tools and platforms of independent MT developers and ‘power users’ through an API that TAUS makes available. Some power users have oceans of data and will not be in need of such a shared pool of resources. Other power users can never have enough translation data. A new dimension in data is the phonetic data. Not so many, but a few of the largest, TAUS members strongly suggest that the Data Cloud expands into collecting and sharing speech data.
This theme links to Machine Learning and Machine Translation.
Connecting tools and exchanging files across different translation and content management systems without loss of leverage and data remains one of the biggest challenges in the translation industry. These mundane problems form a hindrance to advancing automation and innovation in every individual business. Fixing the ‘pipes’ and the ‘connectors’ is a prerequisite in every technology strategy. The last time we counted, there were over sixty different APIs managing the traffic of translations in the industry. One idea submitted was for TAUS to set up an open exchange market for ‘connectors’. In the past five years TAUS has worked on and off on the design of a standard common translation web services API. Last year TAUS shared the specifications and resources with GALA and the GALA organization has now taken the lead with what is now called the TAPICC project. A deliverable from the Industry Summit could be a roadmap for a standard translation API that can help overcome the mundane problems of our industry.
This theme links to Academy.
A vital component in every innovation and automation strategy is the level of skills and adoption in the company’s own organization and in their supply chains. Robots may take on a bigger role and bring efficiencies, the human side of localization remains indispensable. The success of innovative strategies and technologies like crowdsourcing, machine translation, machine learning depends entirely on the acceptance and embracement, we should say, by the people who work with us and for us. Discussions at the Industry Summit should focus on how we, TAUS as an industry organization, can best support its members in driving adoption and preparing a next generation of language workers. Ideas that were submitted mention the push of best practices, providing courses, partnering with academia, giving advice, offering consultancy, webinars and business intelligence reports.
This theme links to all other themes.
The many questions related to the development of innovation strategies will be ‘orchestrated’ in a meeting format known as ‘The World Cafe’ (see: http://www.theworldcafe.com/). The participants will meet in small groups of four (max five) at small cafe style tables to discuss specific questions. There will be progressive rounds of conversations, approximately 20 minutes each. Notes will be taken on the tablecloths, index cards and flip charts. Reporters are appointed to record the collective knowledge as it develops. The small table conversation are alternated with town meeting-style conversations to synthesize the discoveries.
Directors and business owners from the following companies (preliminary list): Alpha CRC, Across, Amazon, berns language consulting, Booking.com, CA Technologies, Capita TI, Daimler, Dell, eBay, GALA, Google, Iconic Translation Machines, Intel, Intuit, John Deere, Kaleidoscope, KantanMT, Karlsruhe Institute of Technology, Lionbridge, Microsoft, Nike, Oracle, PTC, SDL, STP Nordic, tauyou language technology, TNT Express, TransPerfect, VMware, Welocalize, XTM
Organizer: Jaap van der Meer
Host: Colin Gills
Program coordinator: Anne-Maj van der Meer
Reporters: Isabella Massardo, Dace Dzeguze, Anna Samiotou
Top Down, Bottom Up and Right On
The flow of the conversations will go from top to bottom and back up to find exactly the right questions to ask and the best answers to give. See the big picture, how the six innovation themes interconnect and depend on each other. Can one go without the other? What are the inhibitors? What are the drivers?
Zoom in on Machine Translation: what do we need most to make it successful? More data, better education? More advances in technology? What are the opportunities and strengths? What about interoperability? Why is it that we can’t get this fixed? Let’s fix it! Write down the steps and the time line. Who owns the data? Who has enough data and the right data to do what? Train machines! To do what? Algorithmic management. How does that affect the jobs of our people? Quality prediction. What does that mean? Can we eliminate steps in the quality review cycle, once we have enough data? What is our competitive edge in machine learning? How big are the investments we need to make? How do we set prices in the new gig economy?
Zoom out again. How is the future unfolding itself? Or rather: how do we redefine our business? Where is the value add in 2021 of the language service provider? What should translation buyers ask for or wish for? And more to the point now, and working towards concrete deliverables: how can we work together to control our destiny and get the best results for everyone? Can we start an initiative to collect and share speech data? Can we share algorithms for machine learning? What is the most essential business intelligence we need?
In the period from mid February until the Industry Summit we plan to interview many of the participants to better understand their expectations and collect more questions that will help us to focus the discussions.
Visit TAUS head office and ‘pub crawl’ and informal Italian dinner in the Jordaan district.
Boat tour through the Amsterdam canals and dinner overlooking the Amsterdam IJ.
More information and addresses will be published shortly.
The Industry Summit will be held in the city center of Amsterdam at De Rode Hoed. De Rode Hoed is a debat center housed in a former Remonstrant clandestine church. For this special event, TAUS has reserved the Oosterhuiszaal, which is the biggest room of De Rode Hoed and the space that served as the clandestine church.
Address of De Rode Hoed:
1015 CV Amsterdam
TAUS has reserved a block of rooms at the Hoxton Hotel, which is in walking distance of De Rode Hoed. The Hoxton is located in the building where the mayor lived in the 17th century. To book your room at the Hoxton, please contact Anne-Maj van der Meer to receive the information and booking sheet. Update (16 Feb): as of today the Hoxton is fully booked. See below a few other hotel recommendations. All are in walking distance of the venue.
The Times Hotel. Address: Herengracht 135-137, 1015 BG Amsterdam
't Hotel. Address: Leliegracht 18, 1015 DE Amsterdam
Hotel Sebastiens. Address: Keizersgracht 15, 1015 CC Amsterdam
Hotel Estherea. Address: Singel 303-309, 1012 WJ Amsterdam