The question is: are we all on the same evolutionary path? Or do some of us take different turns, make shortcuts and even arrive in different places? The TAUS Executive Forum on April 25 and 26 in Beijing opened new perspectives on translation in China that many of us had perhaps not expected. China typically copies what others have already built and done before them. Fast trains are modeled to the TGV in France, electric cars are inspired by Tesla, and fashion in China follows the trends around the world closely.
So when Eric Liu of LocaTran Translations at the TAUS Forum said that most language service providers in China are not yet using Translation Memory (TM) software, it was implied that we have to wait until they are all on board. But what if perhaps they skip this TM phase altogether and jump right into much more advanced technologies and platforms? What if the magnitude and pressure to translate everything in a market like China are so great that old tools and tricks simply do not suffice?
If we look at what was presented at the TAUS Forum in Beijing last week, we have to admit that western companies do not have exclusivity on innovation in translation. The evolution path TAUS has pictured for fifty years in the translation industry illustrates that we are entering the Convergence Era: technologies, services and business models are converging to form completely new and compelling solutions for companies, governments and citizens who are lost in languages and translation. Pactera, CSOFT, UTH International, GTC Tech, Tmxmall, Lingosail, all Chinese translation companies, showed that they are (getting) ready to meet new demands from increasingly open and globalizing markets. And by the way, translation memory wasn’t mentioned at all, or perhaps just in the sidelines…
Automated Crowdsourcing and Big Translation
Pactera, the largest language service company in China (according to CSA), is betting big time on crowdsourcing translation as the model of the future. In their presentation, at the TAUS Forum in Beijing, Fulin Liao and Jonathan Gao laid out the strategic plan of the company to shift from traditional outsourcing to automated crowdsourcing. Disintermediating the cascaded supply chain, that is so typical for the translation industry, has been the focus of a number of start-ups like Gengo, Say Hello, and AnyDoor. However, for a big global translation company, this seems like a unique and courageous move. In the past year the Pactera developers have been working diligently developing all the required components of the new crowdsourcing platform under the directive of: “We don’t quite care who is doing what, but we do care what is able to be done well by when.”
CSOFT, another large language service provider headquartered in Beijing and best known perhaps for their TermWiki platform, presented their brand new Stepes mobile translation technology. This innovation had already caught attention in the press with an article on Big Translation in Fast Company magazine last January. Carl Yao came to the TAUS Executive Forum in Beijing to demonstrate the patent pending app and to explain the motivation of CSOFT to bet all its cards on mobile translation. There is a convincing and powerful argument that the established workforce of an estimated quarter of a million professional translators is far from sufficient to meet the ever-growing demand of globalizing economies. So why not tap into the potential translation capacity of more than three and a half billion people in the world who speak more than one language. They, as a matter of fact, may often be better suited to take on certain translations because they are subject experts in many domains. All that was needed was to build an app for the phone that makes doing a translation as easy as texting. And that is what Stepes (to be pronounced as steps) is all about.
Big Data and Deep Learning
UTH International from Shanghai — another LSP from China that is likely to become a world player in cross-lingual information services very soon — believes that the official numbers on the size of the translation industry are grossly underestimated. According to Henry Wang, the latent market is much bigger (perhaps ten to a hundred times), especially if we include ‘For Information Purpose Only’ translation needs, and the only way to address this is through Big Data. UTH has already aggregated 4.5 billion clean translation units in different vertical fields and is using this to empower its new Sesame Translate platform as a personalized automatic translation and foreign language reading assistant that will easily beat Baidu Translator and Google Translate.
And then there was Global Tone Communication Technology Company Ltd., or shorter GTC Tech, the company with probably the longest history in translation in China. Originated as a dedicated translation unit for the United Nations in 1973 the group became known as the China Translation & Publishing Corporation (CTPC), and it was the official translation provider for the Olympic Games in China in 2008 and the 2010 Shanghai World Expo. GTC Tech is the holding subsidiary of CTPC and underscore the strong focus of the new management team on technology. Some three hundred engineers are working hard to prove to the world that language barriers really belong to the past. And to say that they are ‘just copying’ Google Translate would not do justice to the impressive range of technology offerings under the Yee brand.
YeeCloud is an O2O ecosystem for all translation needs: human, machine, project management, and community. The YeeCloud Web Translator translates websites on the fly. YeeKit is an open translation toolkit and resource-sharing platform with machine translation available in 32 languages. The average TAUS member and follower has seen all of this before and will probably say: “what’s new?”. But wait until you have seen YeeSight. YeeSight is the cross-language big data analytical platform where machine translation, sentiment analysis and internet search mash up to create a unique service for companies, governments, institutions and anyone who wants to analyze and forecast market trends, and make assessments of economic and financial trends. Imagine that you represent a western car manufacturer and you need to assess and compare your company’s position in Asian markets. YeeSight will offer the intelligence to you in nice graphs instantly or overnight, only leaving you to wonder how in god’s name the technology under the hood figured all of this out from the hundreds of thousands of articles and social media web pages written in so many different languages. Being government owned, GTC Tech must have deep enough pockets (and strong enough supporters) to push the development of all this technology to its limits. And what these limits are, it is hard to tell. Don’t think that deep learning and Neural Machine Translation has not arrived in China yet. IBM’s Watson has some serious challengers here. The management told me that an IPO for GTC Tech is underway.
Data Cloud and super-human MT
Tmxmall is a start-up in China infected with the spirit of the TAUS Data Cloud. The founder Jing Zhang raised 2 million RMB from an angel investor to build a platform that allows language service providers to upload and share translation memories publicly or in private vaults. The goal is to develop a market place where buyers and sellers can trade their translation memories with the aim to reduce translation costs, save time and improve quality. Zhang’s presentation at the TAUS Forum raised questions from the audience that sounded quite familiar to us. How do you manage and secure the quality of the translation memories? What about the copyrights on the TMs? We are confident that Tmxmall will succeed in securing a second round of investment, and we are anxious to follow how the data sharing practices will play out in China.
Long-term TAUS member Lingosail shared its experiences with its own developed hybrid machine translation technology. The CEO Yongpeng Wei presented insightful profiles and feedback from some of his translators on the use of the MT technologies. Without exception, they were positive and felt that the post-editing practice helped to be more productive and more accurate in their translation activity.
What news could the westerners bring to this illustrious group of Chinese innovators? In our audience at the Beijing TAUS Forum, we had Chris Wendt (the leader of Microsoft’s MT group) and Mark Seligman (CEO of Spoken Translation) who both spoke about speech-to-speech translation, its challenges and rapid developments. We also had Jean Senellart (CTO of Systran) demonstrating his enthusiasm over the speed with which we can expect Neural Networks and deep learning improve the performance of MT. He coined the new term ‘super-human MT’ and got the attention of all the academics, LSPs and translation buyers in the room.
Are you curious to see where all this may lead to? Come to the TAUS Executive Forum in China in 2017 and you will be surprised. One thing that might happen then is that WeChat also offers Speech-to-Speech translation and a good interface for crowdsourced translation volunteers. The future is a world without language barriers. How we are getting there is different in different places. China has its own characteristics. The needs are more intense. Access restrictions on some western internet services may in fact create a fertile environment for blue sky thinking and surprising innovation. Indeed, think how WeChat became a universal app for chat, social media, payments and much more (WhatsApp, Facebook, PayPal, all in one). The Belt and Road government-funded program pushes Chinese companies to open up and communicate better with the world. Compare this with the CEF project in Europe that is primarily concerned with the internal language barriers among the member states.
Language business innovation – made in China – is surprisingly interesting.