Five key lessons of 2021, and how a Translation Management System can help you learn them
The 2020s have started with uncertainty. It’s reassuring, then, to have technology we can rely on. In 2021, Translation Management Systems offered security, economy, truth and personalization. In this white paper we look at the lessons of the past year, and the tech that kept users on the front foot.
- Artificial Intelligence – real benefits from vector spaces
- Data driven decision-making
- A single, secure source of truth
- Towards a tech democracy
- Personalization at scale
1. Artificial intelligence – real benefits from vector spaces
The value of AI in global communication is growing, and in 2021 that growth accelerated. Google CEO Sundar Pichai believes AI’s impact on our lives will be greater than that of fire or electricity. In the short term, PwC research revealed that 86% of companies saw 2021 as a watershed year with AI becoming a mainstream technology for them.
In language technology, AI takes its most advanced form in neural machine translation. 2021 saw increasingly useful NMT applications. Helpdesk and customer service operations needing swift, accurate translation boosted response capacity without having to exceed hiring budgets. NMT support chatbots calculated basic caller needs, acting as first line responders, and connected callers to agents who spoke their language and solved their problems. And neural machine translation of helpdesk FAQs enabled greater self-service problem solving.
If this was the application of NMT, where is the innovation? It’s found in the Natural Language Process AI space, of which NMT is a part. After launching Google Translate in 2006, Google’s research steered towards autocomplete predictions, calculating which words should be added to incomplete language segments based on surrounding words. By using two distinct algorithms coupled with a large neural network, precise predictions could be made. Each word was linked with a mathematical vector of 300 values, and the determining technique was named Vector Space.
This Vector Space is a relationship map unique to each language. Connecting these spaces between languages promised to open the door to game-changing insights. XTM picked up the baton from Google in 2020. Our AI team proved that, given enough bilingual data for two languages, Vector Spaces could be “normalized” to create an Inter-language Vector Space (ILVS). More information on this technology is available here.
Applying the ILVS framework has enabled us to build functionality that checks on the accuracy of what other forms of AI – such as NMT – produce.
XTM functionalities fuelled by ILVS enable users to:
- Make more successful bilingual terminology extractions that deliver better terminology translations.
- Align files more precisely, producing larger Translation Memories with more valuable information.
- Get better automatic suggestions for inline element placement. The measured accuracy of auto-inline element placement is 98%.
2021 gave ample proof that NLP AI is the next key field of AI research. NLP AI sessions at multiple industry events, including many of our own, showcased the benefits of a balance between AI, machine learning and NLP.
2. Data-driven decision making – the clarity of the TMS
If a process can’t be analysed it can’t be measured. If it can’t be measured it can’t be valued. In 2021 we saw data-driven initiatives fast-track years of digital transformation into months. A survey by The AI Journal found that 74% of executives expect AI to deliver more efficient processes. 55% believe it will help to create completely new business models.
It’s hardly insightful to state that business decisions are driven by data. Insight is needed, though, in the choice of data to prioritize. This choice can make the user feel like a TV viewer with remote control access to hundreds of channels.
For the user with an eye on global performance, a Translation Management System offers clarity. Tuning into this channel brings detailed reporting on translation memory leveraging, and statistical breakdown per language, project, cost centre, organization and time period. Real-time project analysis, as granular as the user needs it to be, enables continuous process control and process improvement.
By gauging response times and productivity of employees and freelancers, and taking into account satisfaction scores for previous jobs, users can assign tasks with an informed view of their likely completion date.
A Translation Management System empowers the user to build their own bespoke analytics, measure performance, allocate resources and predict outcomes. Centralizing different categories of data in a TMS enables swift cross-referencing. Users needing audio-visual localization, for example, can access files in multiple languages, with data captured on the speed and cost of recording each one.
A TMS enables detailed analysis and maximized ROI for each target language. An overview of vendor MT activity that can focus on any stage of the project process will help users identify the best-performing MT engines, assess post-editing productivity and make informed decisions on future use.
Of course, “data” isn’t limited to the alphanumeric. In the video game universe, gamers who reject a product may be unhappy with the imagery, the translation or even the performance of a voice actor. Analysis of player sentiment in game forums or on social media may give clues that a broad-brush approach misses. Our definition of “data” includes any information, in any format, that can offer insight. Our definition of a Translation Management System is a tool that, among many other things, makes that data as easy as possible to analyze and leverage for the user’s benefit.
Client feedback in 2021 told its own story:
The many benefits realized include customisable workflows, automated cost calculations and the ability for all our linguists and project managers to work from anywhere in the world at any time.“
The problems the TMS solves include enabling us to automate our workflows: we have a direct connection to our development teams; they are now able to send us text through the GIT connector and to create tasks automatically in XTM as soon as they have new or changed texts. We have also saved a lot of money thanks to the TM, since we didn’t benefit from any fuzzy matches in our old proprietary translation tool.
On top of that, we can create our own costs reports at the end of the month and use them for our POs. Costs are calculated automatically based on our agencies’ rate cards, so we only need to extract them from the tool and send them out! In short, it removed a lot of manual steps from our workflows. “
3. A single, secure source of truth – localization technology you can trust and build on
In 2021 the twin threats of cyber security breaches and the spread of misinformation cast a shadow over global communication. In an era of fake news, where can we find objective truth? And in an era of real cyber threats, where can we find reliable data security?
With so many staff switching to remote work, data security needed a fresh look. Homes lack the security of centralized offices, with less secure firewalls and routers, and no onsite IT support. Personal devices used for professional tasks left users open to the risk of data breach.
At the same time, false news reports added to the communication challenge. In the midst of a pandemic, almost 80% of US consumers reported seeing false news items on coronavirus. With unregulated media offering people easy routes to creating and sharing their own versions of the truth, it was easy in 2021 for rumour to become “fact”.
Where can we find truth and data security? The Translation Management System has replaced Localization as the linchpin fifth element in the Globalization Matrix. It’s now a key part of the global content stack.
The language technology provider’s job is to relay data faithfully, ensuring the client’s truth is clear to its target audience. It’s also our job to give the client a truthful picture of project needs. When different LSPs quote for a project, they use their own Translation Memories and tools to calculate weighted and unweighted word counts. Costs and timescales they quote reflect that.
A TMS enables instant, accurate quotes that serve the user, first and last. It’s a single source of truth. It maintains privacy by sharing data on a “need to know” basis. A linguist working in a TMS should be given access to all information relevant to their task, and none that isn’t.
A TMS makes a security promise and keeps it. XTM Cloud uses the secure https protocol. It offers the same level of protection as to secure credit card payments, with translation data safely encoded. No customer has access to data other than their own. Users can be tied to a device, offering further insurance against rogue access.
In 2021, the content of corporate annual reports was even more keenly anticipated than usual. How were industry leaders responding to the year’s social and economic shifts? The answers from board members, in-house counsel and financial officers were crafted in detail. Final drafts represented their truth, and for non-English speaking employees and investors it needed to be shared securely and faithfully. The duty to keep stakeholders informed was honoured, thanks in many cases to the power of centralization and automation.
As one localization professional explained:
Now we have control of our translation memory and assets. Before XTM they were scattered around different language service providers, often inaccessible and rarely secure. It’s the best technology investment we’ve ever made.”
In 2021 users cried out for a single, secure source of truth. XTM Cloud users found it. Truth is a constant, and the language technology used to preserve it is constantly evolving. In 2022 we expect the TMS to keep developing as the key enabler of clear communication. Learn more about why the TMS is the central L10n hub in the technology stack here.
4. Towards a Tech Democracy – choice, motivation and the Translation Management System
Remote work, now a reality for millions, can boost productivity and aid work-life balance. In 2021 XTM’s global growth drove us to hire 78 new employees. All work full or part remotely. Data from workflow tools can tell us who is “performing”, in any location, but spotting patterns and influencing them are two different things. When an employer’s message is shared across territories, its impact must be preserved in each of them.
A Translation Management System will store phrases and terminology central to a brand, save time and money and cut out ambiguity. And company-wide adoption of a TMS can lead naturally to company-wide localization. Teams once low-priority for localization – HR, for example – can have communication brought up to speed in any language.
The employer’s message to new starters should be clear. It will vary between countries, of course, and while the formality of German or Japanese will contrast with the more demonstrative language of southern Europe, a TMS will ensure that each message is localized.
Employee demotivation and disengagement, which rose notably in 2021 amidst the uncertainty of Covid, is a drain on individuals, organizations and national economies. In the United States alone, disengaged employees cost the economy $350 billion per annum in lost productivity.
Smart onboarding is one antidote. Early nerves that come from being unfamiliar with a company’s brand values, processes or tools can be countered. Language units stored in a TMS can be re-used, increasing speed, consistency and quality. A new linguist or project manager working on a first project can be put at ease by using existing assets. They hit the ground running with the confidence and motivation that comes with accrued knowledge.
And the message to established staff?
Is there anything more deflating than seeing days swallowed up by the mundane with no time left for the added-value performance you were hired for? Or more frustrating than avoidable mistakes that undermine hours of good work? People deserve time and freedom to use their skills. They should have safety net protocols to guard against avoidable mistakes. Automation enables more of the good work that lifts employees up and limits the errors that pull them down.
TMS providers should never forget that technology exists to meet user needs. In 2021 XTM created Wish List, a Feature Request Portal that asks users to shape our product. Their feedback is already part of our 2022 roadmap.
Power and choice belong with the user. XTM’s vendor-inclusive TMS has no language service element, so users can choose the LSP they want for any given task. 2021 marked a step forward in users asking for real choice from their language technology. We welcome this trend. As one of our users put it:
My favourite aspect of XTM Cloud is the doors it opens. I saw both sides of the vendor neutrality debate as a Project Manager and for me it’s a no-brainer. When I had a vendor-neutral TMS I had a platform to seek out the best people for the job.”
The employee experience can be enriching no matter where people work from, and a TMS can enhance it every day. That’s tech democracy worth voting for.
5. Personalization at Scale
Media created in languages other than English are in demand. The work has always been good. English speakers are finally catching up with it. In 2019 Parasite, made in Korean, became the first non-English language film to win the Best Picture Oscar.
In 2021 Korean content was in the spotlight again when Squid Game became a hit for Netflix. In their 2021 language report and survey, Duolingo revealed that Korean language learning had risen sharply, coinciding with the soaring popularity of the show.
Good news for linguists? Not entirely. Korean commentators have lamented poor quality subtitling for Squid Game. Key plot and character nuances have been lost, as the streaming boom gives linguists too much to do. The localization industry was built on the assumption of demand for content created in English. In 2021 it felt the weight of demand for content created in other languages.
Personalization at scale is a challenge. With finite human translation resources available, smart automation is helping us meet it.
A Translation Management System gives a cinematic view of project progress. Visual Modes provide visual context, with changes to language segments instantly displayed. Linguists can see subtitles in context, and links from YouTube, Vimeo and other public sites are supported. Users who seek to recreate an immersive audio-visual experience have all the help they need.
Tools that assess MT accuracy enable linguists to deliver a personalized polish to machine translation. It’s the key to localizing the trillion megabytes of data created every day, as well as existing content that global audiences are discovering an appetite for.
One user comment explains how a TMS can aid productivity no matter what the content goals:
XTM gives a clear view of translations in real time and in context, makes linguists more productive, and the automation doesn’t lower the quality of the translation. It actually improves it because translators get to focus on the tasks that need their skills.”
Personalization engages users and creates brand ambassadors. Personalization at scale opens the door to cultural and economic transformation.
In 2021 we emerged, at least partly, from the shadow of COVID-19 and saw the changes it sparked. Digital transformation and the shift towards remote work have altered the way we do business. Throughout the year, AI-powered language technology helped us communicate more swiftly, accurately and widely. 2022 will see further steps down this path, and the Translation Management System will continue to light the way.