The question is not whether new technology will replace humans, it is how these advances can be seamlessly integrated into the workflow — and then how to price them.
According to a Computational Linguistics PhD, pricing “is still unstable” for tasks such as machine translation post-editing (PEMT), while a CTO says there are too many overpriced tools that remain “counterintuitive, counterproductive, and problematic” in practice.
Moreover, sans human intervention, “some tools are still not polished enough to produce an outstanding product,” says a BDD; while a Disney vet says there should be no cause for concern because, while new dubbing technology can do a “read,” it cannot “deliver a performance.”
Rafał Tarwacki – Diuna
Rafał Tarwacki came to translation and localization via the software industry; in particular, through OCR (optical character recognition). “I was interested mainly in OCR mechanisms: techniques for sign and text recognition in graphic files, and the application of such mechanisms in document management systems used by translation agencies,” he told Slator, adding, “I see great potential in microservices and new OCR technologies, and that’s what I’m going to focus on.”
Tarwacki was appointed Chief Technology Officer of Diuna, a language service provider (LSP) based in Warsaw, Poland, back in February. According to Tarwacki, who reports to CEO Piotr Kolasa, “We plan, among other things, to develop the OCR/DTP (desktop publishing) department, to implement and expand the order management system and e-learning, as well as work on proprietary mobile applications.”
Prior to Diuna, Tarwacki worked as Lead Programmer at software company Innotion, where he was mainly responsible for the design of document management systems (DMSs). At Diuna, he intends to implement the Scrum project management methodology, a staple in software development because, he explained, “apart from searching for and designing intuitive tools that help solve our daily problems, work organization itself is equally important.”
Tarwacki said Diuna relies mainly on SDL Trados Studio as a productivity tool, “although we are intensively exploring other, also online, tools. As for the TMS, we are currently using Space TMS, an entirely new tool on the market. It is still very little known, but it outperforms most translation project management systems we have dealt with before.”
Tarwacki pointed out that the market is, at present, “full of solutions that are far over the budget of most translation agencies, which, in practice, often turn out to be counterintuitive, counterproductive, and problematic.” He added that since their goal is expansion and development, “tools that are overpriced or, even more so, tools that are not suited to our needs won’t help us in any way. Therefore, we focus more on creating our own solutions that are tailored to our needs and meet our personal expectations.”
As for their efforts as regards neural machine translation (NMT), Diuna has partnered, according to Tarwacki, “with one of the most renowned linguistic schools in order to include our needs and the specificity of NMT in the syllabus.”
Mike Cardinal – IYUNO
After more than a decade at Disney, Mike Cardinal joined IYUNO as VP of Global Dubbing on April 29, 2019. He is based in London and reports to COO Stella Yoo.
Cardinal, who oversaw the localization of all TV content for children across 22 EMEA markets at Disney, noted how there was a “challenge of maintaining creative quality while timelines were being squeezed and pressures were being placed on costs.”
So in his new role at the Singapore-based media localizer, his plan is to put quality at the center of everything they do. Cardinal is responsible for “all eight of our local owned and operated dubbing studio teams, our centralized mixing team, as well as the operational team that manages all of our global projects — essentially anything that is dubbing-related falls under my remit.”
On whether they are looking at any new markets, Cardinal said, “While they are not considered emerging markets by most, we are keen to expand our owned and operated footprint into Europe and Latin America.” Meanwhile, they “are putting a great deal of effort into the consolidation of their Asian markets on the back of substantial recent studio investment.”
Cardinal told Slator, this studio investment comprises a new studio build in Manila, a studio acquisition in Japan, a new facility/location in Thailand, a refit in Seoul, and a facility/location “coming shortly in Shanghai.”
Asked to comment on new dubbing technology that automatically syncs an actor’s lip movements to a different language, Cardinal described it as exciting and said he is “keen to see” how such technologies “force us to analyze the methods and processes that we use.”
He added, however, that “at the moment, I don’t think there should be a concern that this is going to take over the dubbing industry as we still need an actor to deliver a believable and empathetic performance; and I have not seen any technologies yet that can deliver a ‘performance’ rather than [just] a ‘read’.”
According to Cardinal, he is also, like everyone, fascinated at how AI and machine learning can “both complement and challenge our existing processes — especially around script translation and adaptation.”
“I am also interested in some of the voice print analysis software that is being developed and the possibilities that opens up in terms of casting, and the ability to match a domestic voice against a database of local language voice talents,” he said.
Michael Ollitervo-Murphy – Unbabel
“What we used to think of as an industry vertical is starting to become more matrix-like,” CX expert Michael Ollitervo-Murphy told Slator. Ollitervo-Murphy joined Unbabel as VP of Customer Experience on April 1, 2019, reports to CEO and Co-founder Vasco Pedro, and is based in Lisbon, where the company was founded.
Ollitervo-Murphy explained the evolving language industry vertical thusly: “For example, whilst we see rather low penetration across traditional financial service providers — banks, insurance, investments — we see a much higher adoption rate amongst the next-generation providers, such as Monese.”
As to where their top buyer demand for neural machine translation (NMT) comes from, he said, “Travel is an obvious area of high adoption for us. With numerous global OTA (online travel agents) and sites allowing us to compare before we commit, prices fluctuate, availability of flights and accommodations change constantly; and, in order to maintain profit margin, NMT is the best way to deliver great service at profitable cost.”
Another: online gaming communities. Ollitervo-Murphy described them as “24/7 connected, passionate brand ambassadors, and invariably Gen Y or Z.” He noted how “the old model didn’t fit [because] if you’re needing help in your online game, you can’t wait six weeks for a company to recruit and train a native speaker!” However, the “largest opportunity” for the company still comes from the BPO industry.
Ollitervo-Murphy, who describes his personal approach to CX as “a tad unconventional,” said, “One of the worst kept secrets in the CX and customer service world is that how you treat your own people directly impacts how they will treat your customers. It’s important that both feel respected, valued, empowered, and engaged.”
“My management style is probably not too typical in the industry,” he added. “I utterly believe in trusting my people, so I’ll never put barriers in their way. I am there to free them up to deliver to our customers. So whilst overall I have a very liberal, hands-off, and hierarchy-free style, my lifetime of experience […] has taught me that to get the best of your people, you need to compromise part of your way of thinking and behaving to enable their abilities.”
He said it has less to do with age, background, ethnicity, location (“there are too many variables”), and more to do with focusing on the individual, identifying their “motivator buttons,” and helping them achieve their potential.
“Some people find this a little scary, at first,” he admitted. “But the proof is in the pudding. My teams consistently deliver delight to our customers,” he said.
Carla Parra Escartín – Unbabel
“As someone who has been in the translation industry and machine translation research areas for the past 12 years, it has been interesting to see how there has been an evolution from completely neglecting MT (machine translation) to accepting that MT is here to stay,” said Carla Parra Escartín, who joined Unbabel as Director of Linguistic Services on March, 25, 2019. Escartín speaks six languages and has a PhD in Computational Linguistics. She reports to CTO Joāo Graça.
So complete has the turnaround been toward MT, Escartín pointed out, that it led to “the creation of a new task in translation, namely post-editing (PEMT).” But even as more professional translators embrace PEMT, she said, “what is still unstable in the industry is the pricing system for these tasks.”
Escartín explained that since MT quality is not something known beforehand, “it is hard to assess which discounts to apply to post-editing tasks. Many professional translators compare this situation to the time when Computer-Assisted Translation (CAT) tools disrupted the market and provoked discrepancies in the way contents from translation memories should be charged.”
The big difference this time around, according Escartín, is that “while it can be assumed that the contents of Translation Memories will be usable, and CAT tools will indicate to translators the main differences in the source text so that they can quickly identify what should be fixed, in the case of MT it is the translator who has to first assess the MT output, then think how it can be reused — and, ultimately, post-edit the text.”
“At the same time, research has proven that, when MT is of an appropriate quality, translators are more productive when post-editing and, hence, a discount seems reasonable,” she said, disclosing that they pay their editors according to time spent editing MT. “Hence, if MT is of a worse quality, they don’t get ‘penalized’ in their rates. This is something that I have also heard in other translation circles, but is not widespread yet, and should be integrated in translation workflows including MT in a seamless way,” Escartín said.
“We are also still lacking ways to automatically report MT issues, and post-editing suggestions, similar to what translators observe when editing translation memory content,” she added.
Prior to joining Unbabel, Escartín “worked extensively in the area of Human Factors in Machine Translation.” Her PhD “focused on ways to automatically extract compound dictionaries between German and Spanish to feed Statistical Machine Translation systems and improve their overall quality.”
Since then, the well published Escartín has also worked on the impact of MT on professional translators, how MT can be used by non-native speakers of English to publish their research in English, as well as ethics and MT in crisis scenarios.
So what does she predict will be the next big thing after the now-ubiquitous MT? “If I think of something that is needed in the translation industry, it would be quality estimation integrated with automated post-editing. If I think of what will be the next big thing in MT, it will be something that we are starting to see: speech-to-speech translation in real time.”
Victoria Herrera Mercader – Global Lingo
According to Victoria Herrera Mercader, “Some tools are still not polished enough to produce an outstanding product without human interaction. At the end of the day, translation is an art performed by humans assisted by tools; the better the tool is, the easier the job will become for the human in front of it.”
Mercader, who works as a translator and volunteers as an interpreter, joined Global Lingo (London) as Business Development Director back in March. “I encourage my clients to be open-minded with regards to NMT, MT, ASR (automatic speech recognition) — but when it makes sense,” she added.
She pointed out that the skill sets of professional linguists and the logistics of an interpreting assignment “are usually the reasons why the costs are considerably high.” So embracing “the advantages technology has to offer [can] provide cost-effective solutions.”
As for the consolidation that has swept the industry via M&A, Mercader has this to say from a sales standpoint: “I always see M&As through a positive prism; someone is willing to finish a chapter, while another is willing to continue reading the book. While it is not a concern from the competition standpoint for established companies, it could create a higher barrier to entry. But it is a strategic growth path I respect. For the freelance point of view, negotiating rates might get harder if an LSP has grown and has more work to offer, but this is not necessarily something negative.”
Sheena Makhecha – Anzu Global
Sheena Makhecha took on the role of Globalization Account Manager at Massachusetts-based LSP staffing company Anzu Global on April 8, 2019. She reports to Managing Partner Mike Klinger. Based in California, Makhecha will be responsible for staffing services on the West Coast.
Makhecha is a localization industry veteran of more than a decade. Prior to Anzu, she worked as Account Manager and Project Manager at LSPs such as Lionbridge, TOIN, and TransPerfect.
She said, “One of the challenges I see in my role is this is my first time having a career in Staffing. As this is a new space for me, I understand there will be a bit of a learning curve and ramp-up time.” Hence, she plans to “reach out to the contacts that I have in my network to let them know of my move into this field, and to hopefully be a useful resource in the future for them.”
Makhecha is an active member of the Silicon Valley Chapter of Women in Localization. She enjoys exploring the outdoors with her family, practicing yoga, and running in her spare time.