Machine translation is becoming an entry point to the dynamic world of speech technologies for many businesses across various industries. Earlier this month at SlatorCon Remote, Intento CEO Konstantin Savenkov sat down with Carrie Fischer, Globalization Services Manager at the world’s largest quick-service food chain, Subway. Together, they discussed Subway’s first foray into MT and how it ultimately drove their biggest year in content creation and translation with a quick turnaround time.
Keep reading for key insights from the conversation on MT as a gateway to speech technologies, the natural progression of cognitive AI, and real-world use cases from the foodservice industry and beyond.
• • •
SlatorCon Remote December 2021 brought together Slator’s network of industry leaders and experts to highlight the most promising industry trends and high growth opportunities. A special focus was placed on the speech technologies enabling both providers and enterprises to expand into fresh markets while growing revenues, improving operational efficiencies, and strengthening teams.
Fischer and Savenkov used the platform to highlight strategies for adding Machine Speech to your localization toolkit. More importantly, what makes this a successful endeavor. They reflected on how MT and speech technologies are already delivering outstanding results, and what’s on the horizon for new areas of value creation.
• • •
Here are the major takeaways:
Where we are with speech technologies
We can now see that machine translation tech has passed the level of skepticism that previously hampered adoption. MT is making its entrance into the standard practice for businesses looking to increase productivity, exceeding the brightest expectations of end-users. Many of these companies are beginning to talk publicly about the increases in ROI driven by proper MT. In the case of machine translation, this level of development was relatively easy — facilitated by machine-friendly tools such as TMS/CAT, standard workflows, and trained employees.
Turning towards the broader content creation chain, we see that content authoring, content transformation (between text, images, etc.) are much less developed and utilized. There are fewer end-to-end tools to support the post-editing process, a wide variety of workflows, and not as many trained professionals. With some companies, there are no tools to integrate, with colleagues creating tasks for each other to run on. However, the AI part is already there. The capacity for cutting turnaround times and costs is as great as the success seen by MT. Adding AI to these processes has tremendous potential for generating multilingual content quickly, which is also substantially more accessible.
How Subway learned the benefits of MT
Subway’s experience with MT this past year confirmed the level of triumphant outcomes being discussed. This was Fischer’s first experience with the technology. MT had always been on their radar, but there was never any real need to experiment. That was until Fischer was given a project of about half a million words in English which needed to be translated in a rush timeline of three weeks. By employing MT, the entire project was translated on schedule with impressive results, leaving the localization team at Subway happy with the outcome.
“This has been the biggest year for content creation and translation in a very quick turnaround time”
This experiment spurred what would become the biggest year for Subway in terms of words translated in exceptional turnaround time. A huge tool in the introduction of MT with immediate results was guidance from the Intento 2021 State of Machine Translation report, which provides an evaluation of all need-to-know aspects of the MT landscape.
Speech technologies for e-learning
Later, introducing speech technologies to the mix, Fischer approached e-learning content. Subway required an e-learning course which had to be delivered to franchises. The basis of the program was a cartoon with a voiceover, and they simply didn’t have the budget for the required 400k words in 9 languages. Fischer was skeptical of speech technologies — until she heard the samples.
“In-country reviewers were blown away by the quality of speech”
The need for e-learning content has been boosted by the pandemic. However, we see that the workflows are much less standardized than simple text translation, and there is often no tool to integrate voiceover into the workflow. With the application of robust AI, the complex, scattered workflow can become straightforward, holistically improving productivity. The e-learning example supports the idea that AI-driven applications will make the entire content development chain less fragmented.
Human experts are still needed, as they possess unique knowledge for feedback to improve output. The approach here is not to replace existing vendors. Instead, injecting AI into trusted, established workflows will skyrocket results. Fischer likens the role of Intento as the Wizard of Oz, doing its magic behind the curtain while the client simply experiences the improvements.
The next frontier — content generation
For Fischer’s team at Subway, the next area of focus is the content creation process. Companies in franchising are at an advantage here, because of the centralized aspect of their business. A lot of Subway’s content is informal, teaching ‘sandwich artists’ how to greet guests, make a sandwich, etc. The goal is to make the content as friendly and relatable as possible. Savenkov likens the relationship between human and AI coworkers in generation to teaching a modern-day robot to make sandwiches for the franchises versus the sandwich artists. In the end, you need the human touch and experience, but you need to assist this as much as possible with well-developed AI to ensure universal productivity and quality.