Emerging technologies: Innovations and disruptions

At CSA Research, we devote a significant portion of our time to analyzing the language services and technology market. Our reseach team recently identified trends and technologies that have the most potential for innovation or disruption. We considered a range of technologies for written and spoken language, including the alphabet soup of MT, TM, TMS and IMS, as well as improving translation at the source and dealing with the flood of devices into the market that add some more acronyms to the mix,  such as OTA and IoT.

Machine translation advances written language processing

Machine translation (MT) has been a topic of development and discussion for decades. Big data has accelerated and will continue to drive the evolution of statistical MT, especially now that companies can buy computing power in bulk from suppliers such as Amazon and Google. Deep learning technologies will eventually deconstruct linguistic and semantic components, leveraging that computing power and neural network to improve the quality and usability of the translated content. For example, Microsoft recently released neural net-based translation apps with offline capability for both iOS and Android. While their quality is poor, the fact that they can run at all on limited hardware is an impressive achievement.

We expect to see many more companies embed MT for both written and spoken language, providing real-time translation wherever content is exchanged. They will deploy the technology to meet the classic triad of requirements for more volume, faster turnaround time, and lower cost — in other words, where humans can’t scale. But in the final analysis, humans will be the teachers of machine translation and judges of its quality and usefulness.

In contrast, core translation memory (TM) technology has been quite stable for a number of years, but new entrants such as Lilt and MateCat combine TM with MT in innovative ways. In particular, the boundaries between subsegment TM and statistical MT are increasingly blurry. Predictive (look-ahead) analysis can help locate relevant phrases to assist translators to boost productivity in a more natural and ergonomic environment. Terminology management and leverage will become better integrated in such solutions.

Shared TM has re-emerged as a hot issue. Despite expectations that it would revolutionize the translation industry, previous solutions were too tied to local servers, too complex or had too many legal issues for commercial users. But computer-aided translation (CAT) tools that include sharing options are helping take TM sharing from a niche solution to a mainstream option. Cloud-based solutions naturally lend themselves to sharing in a way that desktop and older client-server architectures did not, so this trend favors products from companies such as Memsource, Smartling and XTM Cloud — and will push traditional suppliers such as SDL and STAR to the cloud. Meanwhile, Language Terminal from Kilgray allows sharing of other resources, such as filter configurations and segmentation rules.

Management systems progress to the next level

On the project management front, automation is greatly increasing and gradually replacing humans on high-volume repetitive projects. Providers such as Elanex and Straker Translations integrate artificial intelligence to their translation management systems and can handle projects from quoting to invoicing without human intervention. XTRF introduced Smart Connectors and Smart Projects earlier this year, making automation more accessible to language service providers (LSPs) without big development budgets. We expect to see an increase of such deployments because reducing internal labor cost is an essential component of LSPs’ competitiveness in the midst of the price pressure they face. This expert system feature is also increasingly popular for high-volume, on-site interpreting assignments, in which fewer human touches for linguist selection and booking enable faster scheduling and lower costs.

Connectivity is no longer an afterthought in buy-side content environments or in LSP operations. Technology decision-makers seek systems with complete application programming interfaces (APIs) or software development kits, as well as connectors for multiple content applications. Translation management system (TMS) offerings are expected to connect to just about any digital source, including  MT software; web content management system software such as Adobe Experience Manager and Sitecore; and file sharing applications such as Box and Dropbox. Competition among TMS suppliers has gone into overdrive to build more and better connectors to make disparate solutions work together, and even developers with solid APIs — such as Plunet and SDL — are retooling their architectures to meet demand for more connectivity and simpler integration points. This trend will continue as both end-buyers and LSPs find the need to connect to just about anything.

We see similar energy and innovation among spoken-language technology developers. Interpreting is vibrant with new developments. With new options such as Cloud Interpreter and Interprefy, delivery is increasingly visual through video-mediated solutions as a substitute for conference, on-site, or even telephone interpreting. Smart providers like LanguageLine Solutions also seek to break the silos that separate verbal from written language services. Their goal is to create ecosystems where buyers can integrate their internal resources, supplement them with video or telephone providers, and even submit written-language translation jobs, all within the same online portal or management system.

We have long seen innovators chipping away at the divide between translation and interpretation, and that delineation decreases every year. Multilingual chat and real-time translation is often powered by interpreters rather than translators. Machine interpretation benefits from MT innovation in systems such as Lexifone, Lionbridge’s GeoFluent, and Skype Translator. Mobile translation apps from Baidu, iTranslate and Microsoft will play a stopgap role until developers begin integrating spoken-language MT capability in those devices.

The evolution of content brings up new focus areas

Translation or interpreting output, regardless of its human or machine provenance, can only be so good, so both end-buyers and LSPs are aware of the need to improve source context. Automatic content enrichment (ACE) is an emerging area of interest. Solutions such as Acrolinx and Open Calais are achieving increased traction in the space. ACE applies natural language processing to add links and metadata to the source or target text to make information more accessible and useful. ACE, combined with human curation, allows localization experts to add relevant information and links not found in the source to particular target language versions. Localizers have just started to explore the potential. The European Union’s FREME project is developing an open framework to support multilingual ACE services alongside services such as machine translation and content internationalization.

Demand for language technology will get a boost from the huge volume of new devices coming online. Mobile device manufacturers predict that 90% of the world’s population over six years old will have a mobile phone by 2020, and the Internet of Things (IoT) will add billions more devices to the environment, each with the potential to connect and communicate in other languages. A lot of localization will be required to intelligently interact with all of the newest wearables and other gadgets connecting to the IoT around the world. Right now, this localization workflow, especially testing, often takes place outside of a company’s standard processes. Products such as Applanga, Qordoba and Smartling’s Jargon already enable translators and reviewers to edit apps in-context and allow their customers to offer over-the-air (OTA) updates to skip app store submission. Users around the world gain immediate access to the latest localized versions. Technology providers are bound to integrate these capabilities into translation management systems, along with more options for browser-based testing for the many flavors of Android and iOS available worldwide.

The supply chain is also ripe for disruption

Just as Uber and other gig economy companies seek to upend the labor status quo by contracting with independent workers short-term, many LSPs are trying to change the composition of their supply chains. Interpreting technology providers such as Stratus lead the way with products to find local interpreters ready to jump on an assignment, while translation-centered LSPs such as CSOFT aim to empower just about anyone to become a translator on their mobile phones — in CSOFT’s case, with the Stepes tool.

What’s an LSP to do about all this?

With all this activity, is the industry really changing? For some LSPs, it is still business as usual and these innovations are far from representative of their day-to-day operations. But those that seek sustainable growth won’t postpone integrating disruptive options to their offerings. Our primary research over the years has shown that only those suppliers that embrace technology can scale to meet their clients’ requirements, grow their business and stay profitable. We’ve also seen failures when LSPs invested too much in development — being a fast follower is more realistic to many LSPs. But ones that do have the vision and capability certainly stand out and can achieve faster growth than peers. Others should stay on a safer stance, but be ready to adopt the innovator models before it’s too late and they can’t catch up.