In this year’s edition of the Nimdzi Language Technology Atlas, we collected data from providers of more than 700 technology solutions.
The data gathering behind the Atlas is based on four main sources:
These sources have given us a comprehensive understanding of the current state of technology development in the industry which we are eager to share.
As compared to the 2019 version of our Atlas, where the number of individual products mapped had increased from 400 to over 500, in 2020 the number grew more than 20 percent to reach 660 individual products. Due to this increase, we had to make the difficult decision to remove smaller technology groupings such as Translation Memory editors and tools for internationalization from the infographics. The current Atlas represents the following nine major groupings of tools:
Let’s have a look at what has changed in language technology over the past year.
Migration to the cloud actively continues. According to the Nimdzi language technology survey, the majority of the tools are now on the cloud:
The MT sector has seen new features such as MT autoselect and new ways of leveraging MT output to predict quality. Together with industry leaders from the buyer side, we expect more MT quality evaluation tools to appear both stand-alone and built into TMS.
Another sector experiencing growth is Automatic Speech Recognition (ASR). That is why we dedicated a separate section of the Atlas to this type of technology. Even cheap speech recognition software offers greater productivity than doing things manually. We have spotted an increase in services offering automatic transcription and synchronization for subtitles such as Happyscribe, Limecraft, Trint, and Rev.
Automatic subtitles are already replacing humans, and machine voiceover is on the rise as well. Even though it is not yet close to becoming what could be called full machine dubbing, interesting developments are being made in the field of voice synthesis. The current automated technology is mainly used for voiceover recordings such as for documentaries, e-learning, news, corporate videos, and audio description.
Dubbing had to go remote due to pandemic restrictions. Remote dubbing has been around for a while, but the industry seemed reluctant to adopt it due to quality concerns. However, COVID-19 has sped up the adoption process, forcing studios to test out and try different solutions. With the lockdown restrictions, most studios in affected countries were shut down, causing significant disruption in dubbing operations. Most media localization companies needed to adapt their operations very quickly to deal with this situation. To do so, they have put in place different remote dubbing solutions. Some have developed their own platforms (because they were already planning to do so before the pandemic) and others have found an existing solution that could be adapted to their current workflows.
Another type of solution that went remote is Virtual Interpreting. As we discuss further in the VIT section, remote interpreting is one of the main language technology advancements of 2020. The corresponding Atlas section on interpreting systems features over one hundred tools.
In addition to going remote, an interesting tendency has been observed in the field of interoperability. In particular, the industry is seeing growth in the number of integrations between providers of different types of technology, and the middleware sections of the Atlas have been developing quickly.
In view of this, industry standards devoted to the topic of integration are gaining more attention—in particular, the Translation API Cases and Classes Initiative (TAPICC), which aims to advance API standards for multilingual content delivery. Companies such as Straker, Lingo24, Lionbridge, Languagewire, and others have adopted it.
Speaking of APIs, the number of frequently supported APIs also grew from just two (REST, SOAP) to over five. However, the lack of a common API across various systems makes it more difficult for the different tools to connect to one another.
Communication between technologies is very much needed since the number of companies developing new solutions is increasing rapidly. For example, last year we had only 20 Translation Business Management Systems (BMS) in the Atlas, and this time there are 33 such solutions.
This type of technology continues to be developed both:
What’s new here is that some of the BMS tools are becoming available for free, for example, Rulingo. In addition to that, many TMS providers such as Smartcat and Wordbee have extensive business management functions incorporated.
Advances have also been made in the largest section of the Atlas, Translation Management Systems. Continuous localization (CL) remains the buzzword here. Some companies have been following this trend, advertising their delivery methods as continuous but not actually providing it.
The demand for CL will continue to grow, but it will need to bypass the limitations of current file exchange connectors. One of the CL providers, Transifex, names an open-source software development kit (SDK) to detect strings and fileless transfer of strings via API as examples of the top features in their development roadmap.
The positive effect that COVID-19 has had on the remote interpreting market is undeniable. Some VIT companies have seen as much as a 250 percent increase in inbound inquiries since March 2020. In the healthcare sectors, requests for both Over-the-Phone Interpretation (OPI) and Video Remote Interpreting (VRI) have increased by more than 50 percent. This is particularly due to the surge in telemedicine. As healthcare providers across the globe are urging patients to call before making an in-person visit to doctors’ offices, clinics, and hospitals, OPI is being used as a means to support people from all language backgrounds. In addition, legal firms have started utilizing virtual interpreting in the form of VRI calls.
Due to worldwide travel bans, interpreting for onsite events has been obliterated. In turn, events, as well as interpreting services to support them were moved online. Given the spike in requests, more resources are needed to accommodate the additional training and setup requirements for new clients, and to onboard and train new interpreters. VIT provider KUDO, for example, doubled its internal staff in just one month. The number of KUDO certified interpreters also increased by 2,000 to a total of 3,500 in the same period. So while the global pandemic has hit the onsite interpreting business hard, it has accelerated the growth of virtual interpreting.
Among the current challenges which slightly decrease the speed of market penetration for this type of technology, VIT providers named the following:
Media localization companies have been proactive and quick in adapting and reinventing themselves to keep delivering multilingual recordings online while respecting quality standards.
Although for many companies this has been a disaster recovery solution and they are looking forward to going back to the studios, remote dubbing technology has proven to be useful for similar situations (a new wave of COVID, earthquakes, floods, etc.) or other more creative ways of applying this technology (for example, a voice talent who’s on vacation and is needed to record a few lines, hybrid solutions, increase dubbing capacity for peak seasons, etc.).
The main challenges that dubbing companies have faced on the remote path are:
Seeing the demand in supporting technology in audiovisual translation, TMS systems continue to add subtitle editing functionality and video previewing to their translation environments. That is why we included such companies as memoq, SDL, Smartcat and Wordbee in this section—even though those are not subtitling editors, but rather tools with plugins and environments for subtitle localization.
Another important note needs to be made about automatic captions: several applications such as AutoCap, Clips by iPhone, Quicc, and a few others added to this category are designed to caption social media content (for example, videos for Instagram, Facebook, Snapchat and the like).
Speech recognition capabilities are quickly being added to many tools such as online meetings platforms. Recent developments in this area include Rev Live Captions for Zoom or Transcribe for Android by Google, both appearing in Spring 2020.
An important—albeit not new—niche that language services buyers are still struggling with is in-context review. In-context review tools help identify errors in translated content and fix them on-the-go right in the working environment. One of the most popular use cases for this is an online tool enabling in-context translation and in-country review of web-based products, desktop apps, and websites.
Even though there are proven solutions on the market such as InContext Translation and QA by Lingoport, Rigi.io, or visualReview by translate5 that help with in-country review, marketing and localization teams across the globe continue to call out this review activity as an ongoing issue. That may be the reason why some LSPs have been developing their own solutions, e.g. InView for Product by Venga Global.
On paper, TMS stands for Translation Management System. However, the localization industry is in disagreement as to what features and functionalities are covered by this name. This is the most burgeoning segment of the language technology market.
The TMS section of the Nimdzi Language Technology Atlas features over 150 technology solutions. The TMS category accounted for 43.75 percent of the responses to Nimdzi Language Technology Atlas survey.
Before the TMS, there were CAT tools. A CAT (Computer-Assisted or Computer-Aided Translation) tool is software that allows a user to work with bilingual text—the source and the target (translation). Its core components include a translation memory (TM), a termbase (TB), and (sometimes but not necessarily) machine translation options. There’s also usually a built-in quality assurance (QA) feature included.
Over time, to get the translation job done faster and make it more scalable, those components were no longer enough. That’s how a variety of management features appeared in a CAT environment, resulting in the birth of TMS. For easier and quicker performance, many CAT tools emerged on the cloud, too. They are the essence of a TMS.
Let’s have a quick look at the history of this technology and the lineage of the oldest tools—from the 1980s to 2010.
|1984||The ALPS computer-assisted translation system emerged in the Department of Language Studies in the Coventry Lanchester Polytechnic University in Warwickshire, UK.|
|1986||STAR Transit: its beginnings date back to the year 1986. Then, in 1991, Transit 1.0 (a 32-bit DOS version) was one of the first translator-dedicated workstations to appear. The concept behind it was different from the other (CAT) tools that were launched on the market during that period. Transit was based on different corpora that a user fed into a database, while the CAT system was a one-on-one segmental correspondence between source and target texts.|
|Late 80s||Trados GmbH began developing translation software in the late 80s. In the early 90s, they released the first Windows versions of two of the suite's major components, MultiTerm in 1992, and Translator's Workbench in 1994.|
In 1997, the company received a major boost when Microsoft decided to use Trados for its internal localization needs. In 2005, Trados was acquired by SDL. This purchase continued a round of consolidation in the language services and tools industry which had already begun to take shape.
|1992||GlobalWare introduced XL8 Code (DOS) - an interactive tool for translating and maintaining the user interface text strings in C language source code. Its strength was in reusing previous translations and automatically handling revisions; its fuzzy matching algorithms scanned new source language files for previously translated materials and made substitutions. Sounds like a translation memory!|
XL/8 supported character sets from other platforms, including Windows, Macintosh, and Unix. In 1994, GlobalWare also brought out complementary packages called XL8 Help and XL8 Documentation.
|1993||Atril’s Déjà Vu was developed as an outcome of experimenting with MT (Machine Translation) systems and TM tools. A preliminary version was finished in June 1993, and the translation tool for Windows 3.1 was born. An interface that integrated with Microsoft Word for Windows 2.0, creating Version 1 of Déjà Vu, was released to the public in November 1993.|
|1994||MultiTrans (originally MultiCorpora) was founded in Ottawa, Ontario, Canada.|
|mid-1990s||IBM Translation Manager (normally referred to as TM/2) was launched.|
|1998||Beijing Yaxinchcng Software Technology Co. Ltd. was set up as a developer of translation software, becoming the first CAT-tool in China.|
At the end of the year, SDL released SDLX, a suite of translation memory tools. SDLX was developed and used in-house at SDL at first.
|1999||The original Wordfast product, now called Wordfast Classic, was developed by Yves Champollion as a more affordable alternative to Trados.|
SJTU Sunway Software Industry acquired Yaxin CAT from Beijing YaxinChcng Software Technology Co. and released Yaxin CAT v1.0.
|2002||XTM was founded by Bob Willans and Andrzej Zydron "much earlier than other Cloud-based TMS".|
In Canada, MultiTrans 3 was released, introducing a new Advanced Leveraging Translation Memory (ALTM).
Huajian IAT was released in China.
|2003||GlobalLink (initially developed as a support tool at eTranslate in 1999) saw large investments after the parent company was acquired by TransPerfect.|
|2004||memoQ (Kilgray) was built in Hungary by three language technologists: Balázs Kis, István Lengyel, and Gábor Ugray. The name of the company, Kilgray, was derived from the last names of its founders. The first memoQ was released in 2006, and it quickly began to compete with SDL Trados.|
|2005||Across Systems GmbH was founded.|
|2006||Lingotek was founded by Jeffrey Labrum in Draper, Utah, United States.|
|2008||Wordbee was founded by José Vega and Stephan Böhmig.|
|2009||Founded by Sven C. Andrä as Andrä AG in 1999, ONTRAM became available as a commercial product in 2009.|
Smartling was founded in New York City by Jack Welde and Andrey Akselrod.
|The core of IBM TM2 was made available to the open-source community in 2010. (TM2 stands for TranslationManager2. It originates from the IBM Translation Manager).|
Memsource emerged as a tool targeting LSPs, with easy access to web applications, lower initial cost, and free licenses for translators.
After 2010, the sector began to develop by leaps and bounds, with tens of TMS being pushed to the market yearly. Since that time, a new TMS has come out on average once every month.
Together with the leaders from the enterprise-level buyer-side, further along the line in 2020 we expect more:
The main issue with having too many tools on the market is the difficulty in integrating them or switching between one tool and another (leverage loss, migration costs, customization costs). For large organizations, this is further exacerbated by the sheer number of legacy tools and content amassed over decades. While they may be keeping an eye out on the market, the weight of their existing operations actively dissuades them from even contemplating a switch in technology. Initiatives with industry standardization may help in this regard, but only in the longer term and with the emergence of an increased number of adopters of these standards. We don’t expect this interoperability issue to be fully resolved within 2020.
The infographic from this research is publicly available and can be reused with the source quoted. It is updated on an annual basis as information about new tools becomes available (or more frequently depending on the pace of innovation).
What is language technology? Nimdzi has organized a series of panel discussions to cover some of our favorite topics in the space.
On June 10, 2020, we published our Nimdzi Language Technology Atlas, the comprehensive resource that maps hundreds of language technology solutions from all around the world. Two months later, after receiving and reviewing feedback from more than three dozen companies who submitted requests to add new tools or change their categorization, we released an update to the infographic on August 27.
Some machine translation providers are holding out hope for MT systems that adapt to document context. Could this development eliminate the need for custom MT engines? Will context-enabled MT help MT achieve human parity? Will we still need to customize a few years from now? Let’s discuss further.