By Bridget Hylak, Asst. Administrator, ATA Language Technology Division
Moderators: Bridget Hylak and Daniel Sebesta
On October 16, 2023, ATA TEKTalks wrapped up its second year with a highly interactive and favorably-reviewed chat with Gabriel Fairman, award-winning CEO of Bureau Works. Audience reactions to this particular episode, which for the first time in the TEKTalks series allowed a short demo of the tool being featured, were overwhelmingly favorable, as participants chimed in everything from “Impressive!” to “A very cool tool, a natural fit for translators already working with CAT tools!” to “Love the very helpful and informative ATA webinar!”, to “Can’t wait to try it out!”
In fact, ATA TEKTalks 2.4 elicited such audience interest that, in an innovative twist on traditional webinar recaps, let’s plunge directly into the audience’s vibrant Q&A. This approach will hopefully immediately immerse readers into the interactive and dynamic essence of this particular episode of TEKTalks, providing a unique perspective on the session’s insights and an immediate taste of the community’s pulse.
Audience participant Richard inquired about the compatibility of Bureau Works’ business management tools with standard accounting platforms like QuickBooks. Gabriel revealed that while BWX natively integrates with Xero, a QuickBooks integration can be created using their API, showcasing BWX’s flexibility.
Patricia questioned the learning capability of BWX, specifically whether it adapts from user corrections. The answer was a resounding yes, with BWX not only learning from these modifications but also attempting to grasp their semantic implications for future sentences.
Another audience member inquired, “Does BWX require one to work online?” to which Fairman replied, “Yes, good question. First of all, well even when I’m on planes, typically I have some access to internet, but the way BWX was engineered is a hundred percent cloud-based. All the services that we rely on are also cloud-based.” His answer reflected the current trend in most language technology tools.
Audience member Candace asked, “What is the learning curve for using BWX? Is it intuitive for someone who has limited experience with other CAT tools?” Fairman’s response was overwhelmingly encouraging. “I think if you are experienced with other CAT tools, I would say the learning curve is very small. It’s more about the practice than the learning curve. You figure it out very quickly. I think one of the interesting examples is I have is just my mother who was a translator, but never used a CAT tool, they were all too complicated for her. She’s already in her seventies and she spent all her life translating in a given way, she had an entire methodology for how she managed glossaries and where she stored information and how she cross-referenced information. So first, it took her a while to wrap her head around even the idea that maybe there’s a better way to work. And then learning the tool itself was fairly quick. I would say within 90 days she was feeling comfortable working with it, and I would say it took her an additional 90 days to be like, I’m never going back…”
Another audience member asked, “Is the AI access through open AI’s GPT-3.5 or GPT 4?” Fairman replied, “We use both. So we have a very complex machine learning layer in BWX, so we’re never training an instance. We’re never trained any instance of OpenAI or Azure. Number one, all the knowledge management resides in BWX. Because we optimize for costs, we run certain computational processes through 3.5, others through four depending on the nature of the process. So just to give an example, tag fixing, we typically run through 4 because it’s a more complex process. Terminology management, we typically run through 3.5 because it’s a simpler process.”
To the audience member who questioned, “How well does BWX support Asian languages?”, Fairman confidently replied, “Beautifully! It performs very similarly across languages, high resource languages, low resource languages, left to right languages, right to left languages, Asian languages. Because the framework is very universal, it’s very different from, let’s say machine translation where you see a very big difference according to language and domain.”
The session also explored BWX’s compatibility with existing translation memories from other CAT tools, confirming that as long as they are in TMX format, they can be seamlessly imported, as mentioned by Claudio.
“We leverage technology so that human beings can be more productive with higher quality outputs, but most importantly, a higher level of authorship over what they’re doing.”
In an encouraging nod to one audience member who admitted, “AI is making me nervous, especially when it comes to translation. Every time I apply for a translation position, the first thing they tell me is that they depend on AI translation. What should I do? Is AI really better than humans?” Fairman replied, “I would deconstruct that great question. I don’t think AI is better or worse. It’s an aid, and I think Bridget described it really well in a webinar that we hosted recently where she compared it to robotic assisted surgery. That’s kind of how we see it. It’s how do we really leverage technology so that human beings can be more productive with higher quality outputs, but most importantly, a higher level of authorship over what they’re doing, right? If we’re spending less time fixing tags and spelling mistakes and making sure that the meaning was carried over and more time authoring the text, that makes us more important. So I’m not particularly scared about AI. I do think that resisting it is not a good idea. I think that embracing it will A, increase your productivity right now and B, it’s going to get you accustomed to a different way of working with these engines. And C, it’s going to prepare you for the future before it starts. I mean, the tendency is that this will be this kind of approach, not necessarily ours, but using some level of AI will prevail, right? And if you have a way already, if you’re starting early, you’re more prepared, it won’t catch you off guard and you’ll be already flying and producing a lot by the time the market catches up with you.
These vibrant and honest exchanges reflected the pulse of the webinar itself. Fairman’s participation in ATA TEKTalks was a deep dive into the groundbreaking language technology offered by Bureau Works (BWX). From the outset, it was clear that BWX isn’t just another tool in the translator’s kit—it’s a “paradigm shift,” according to Fairman, in how translation and localization work is done. Fairman’s perspective, deeply influenced by his own background and his mother’s experiences as a translator, adds a personal touch to BWX’s development narrative. This personal investment in BWX’s success resonated throughout the presentation, highlighting the tool’s design philosophy: enhancing the human aspect of translation through sophisticated technology.
One of the most captivating parts of the session was the demonstration of BWX’s unique features, such as its dialogical engine and semantic QA verifier, dubbed “translation smells.” These aren’t just technical advancements; they’re a new way of thinking about translation workflow, where the tool learns from the translator in real-time, enhancing both productivity and quality.
The term “language flow architect,” introduced by Fairman, encapsulates this new role of translators in the age of AI. This concept isn’t just about managing texts; it’s about crafting the very flow of language itself, blending creativity with technology. This reimagined role represents a significant leap forward, acknowledging that while AI can take on repetitive tasks, the nuanced art of translation remains firmly in the domain of human expertise.
Fairman and the whole BWX team’s commitment to making the tool accessible to a wide audience is evident in its pricing model. At $9 a month, it’s positioned as an affordable yet powerful tool, with the potential to democratize access to cutting-edge translation technology. This approach could potentially shift market dynamics, enabling individual translators and smaller agencies to compete on a more level playing field with larger organizations.
Furthermore, the seminar touched upon the critical issue of data privacy in the age of AI, offering reassurance about the measures in place to protect user data within BWX. This is a pivotal concern for many in the translation community, providers and clients alike, and Fairman’s detailed explanation provided a solid foundation for trust in BWX’s commitment to privacy and security.
In conclusion, ATA TEKTalks 2.4 with Gabriel Fairman was not just informative; it was a glimpse into the future of translation and localization work. Bureau Works, under Fairman’s leadership, is not just developing a tool; it’s fostering a new ecosystem where technology elevates human expertise, creating a world where translators are not replaced but enhanced. As the industry moves forward, the concepts and technologies introduced by Fairman and BWX will undoubtedly play a crucial role in shaping the future of language services.
Thanks to Gabriel and the entire BWX family for visiting ATA TEKTalks!
**
Please join us on Tuesday, March 19th, as we host ATA TEKTalks 3.1, featuring Sara Basile, product director at XTM, a comprehensive AI-powered CAT and TMS tool. Future installments this year will feature Veronica Hylak of Metalinguist (Q2), Boostlingo (Q3), and Philippe Mercier of Wordscope (Q4).
The Language Technology Division appreciates your input and suggestions on any tools, questions or conundrums you’d like to see featured in an upcoming ATA TT seminar – please reach out to us with your suggestions, and we hope to see you soon!
NOTE: LTD TEKTalks are not language tool trainings (there are enough of those available, just Google away!), but rather, TEKTalks are intended to showcase language industry tools and technologies, and especially the people and philosophies behind them.