Mass digitization across the landscape has changed the way most businesses approach MT, with new stock models, progressive innovations in speech technologies, and more customization options available within Enterprise. In light of these developments, organizations looking to get in front of the curve are asking key questions, such as: ‘How should you build your MT strategy to gain momentum in your organization?’, What should be your priorities to take and keep a value-based approach with MT?’, and How can you find your way in the MT world to make the best decisions?’.
The LT-Innovate Value Cast brought together the expertise of Janice Campbell (Senior Program Manager — MT, Adobe), Yulia Akhulkova (Localization, Nimdzi Insights), Michel Lopez (CEO, e2f), and Konstantin Savenkov (CEO, Intento) to answer these critical questions, uncovering the value that machine translation brings to the table for your business — no matter the size or domain. They tapped into available options for stock models and how companies should handle specific cases where customization is not available, based on the recent Intento 2021 State of Machine Translation report. They specifically explored how this data is relevant for small to medium-sized enterprises with limited resources for customization.
• • •
Here are the major takeaways:
The value of MT for the enterprise
Adobe
In the case of Adobe, machine translation was integral in creating an ecosystem that powers multilingual content delivery across all channels so that all customers, no matter their language experience, will have the same benefits. NMT adoption is becoming the go-to tool for quick and efficient universality. In Adobe-authored product content, custom engines have proven successful, and on-demand translations between any given language pair via stock engines are used for user-generated content, such as threads and messages.
Measuring the health of the system, quality of output, customer satisfaction, and costs / ROI to make data-driven decisions. In an 18 month survey of customer usage for two key international markets (Brazil and Korea), Adobe found that there was no major difference in bounce rate, exit rate, time spent on a page, etc., and an actual uplift in app downloads. Very few customers decided to switch back to English (only 6% of visitors), which speaks to the high levels of satisfaction.
Nimdzi Insights
Nimdzi’s data-driven studies from 2021 centering around the value of MT have uncovered a few different key insights. In the Language Service Provider market, MT is becoming increasingly popular, now the second most widely offered service. Nearly 72% of the LSPs surveyed by Nimdzi offered MT services. In turn, adoption is growing, evidenced by 54% of buyers leveraging MT by the end of 2021.
The Nimdzi team also found that in regards to the raw MT output relating to the intended purpose/use of the MT, most users were satisfied with the results, saying that it could actually be used with minimal human post-editing hassle.
“Solutions that would pay attention to the context, perform all evaluations and measurements automatically, and present users with the most suitable MT options are the most desirable way forward for enterprises” (Yulia Akhulkova, Nimdzi Insights)
Akhulkova points to Intento as a central leader in this charge, placing it among the companies to keep an eye on in order to stay on top of the fragmented MT market. Citing the annual State of Machine Translation report, Akulova recommends using it to grow your MT program in the following ways:
- Keep up with new developments in the dynamic MT market
- See what language combinations are supported by 29 MT engines
- Check out which MT engines are performing best for your specific domain
- Gain an understanding of the dataset procedure
- Get an idea of MT affairs with evaluation metrics and scores
- Keep track of open-source MT engine performance
- View data on how custom MT is different from stock MT
Why businesses need a data-driven approach
MT is the best tool to do more with the budget you have, reaching markets that were previously out of reach. Companies have realized that they need to do roughly 20x more translations than they do today with human services. These additional levels of outreach come from use-cases where human translation is not possible, for example, live chats, but also by enhancing existing internal workflows with MT tools. The value of localization isn’t just restricted to the monetary focus — it’s becoming more about accessibility.
MT can help achieve 20x more without increasing your budget 20x mainly by introducing post-editing processes that minimize the time and effort needed for human interaction. This heightened level of productivity opens the door to different use cases, expanding the value of MT even further.
You need to be data-driven internally in order to figure out which MT service provider to use for each project. There is not going to be one single best-fit provider or model across the board in order to make an MT program work continuously. Different approaches will end up being better than others for each project or business context. This will often require a combination of providers in order to get the best possible results. Paying attention to the annual Intento report will help save time by not having to go through this research each time you approach a project.
Data from human feedback is also central to understanding what works best for a unique context, which is why any MT program should be a continuous effort — there is no silver bullet that will make your efforts perfect out of the gate without this data. There is no way to improve without paying attention to user satisfaction. For example, feedback data can help define the monetary value of your MT, in order to compare it with other translation outcomes. You need feedback to get a proper grasp on how costs or productivity are increasing due to fewer post-editing efforts. Feedback highlights the value of your MT. That’s how you’ll ultimately save a lot of money in the future — by making sure you can understand the value of the data today.
