When Ofer Tirosh set out to create a translation company some two decades ago, he was not aware that he had embarked on building a network of freelance linguists. He was aiming to provide much-needed professional services for the rapidly growing tech industry in his native Israel, a national juggernaut often dubbed the Startup Nation. We spoke with him about how his founding of the translation company Tomedes led to the creation of a network of several thousand professional translators, editors, and proofreaders translating between hundreds of language pairs on demand.
- 1 Ofer, when did you realize you were building an expert network for your translation company?
- 2 So your translation company manages a human database as its core asset?
- 3 Ofer, what are the risks and rewards of building a translation company this way?
- 4 Has your translation company adopted AI software in addition to human expertise?
Ofer, when did you realize you were building an expert network for your translation company?
I had knowledge of expert systems from earlier in my career. Expert systems are software applications which codify rules and practices so that computers can “think” in ways that resemble human experts. Of course, the software is not thinking at all but executing code instructions according to a rule base compiled by programmers. It soon became clear to me that this would not fly for a translation company. The rules of language were far too complex and there were too many subtleties to rely mostly on machines.
Even at that stage, almost all the big tech firms were investing heavily in machine translation research and development. Google and Microsoft led the way, but Facebook, Apple and Amazon were also keenly engaged in seeing how translation could be automated. But the task initially was too complex even for the world’s biggest and richest companies to tackle.
So your translation company manages a human database as its core asset?
At the outset, I built Tomedes as a translation company that focused on human expertise in translating from one language to another. We began recruiting linguists who could translate into the languages that tech companies needed to open their products to multinational marketing and sales. We trained these linguists and created certification protocols to ensure quality standards. We also collected detailed information about our linguists, both employees and freelancers, not only about their language skills but about their availability, their academic training, the feedback on their work, and much more.
Eventually, we had assembled a huge base of personal and professional data that helped us match personnel to projects in an exceptionally efficient manner. After a few years, we realized that we essentially were managing an expert network of language professionals. That human database would be essential in scaling our business.
Ofer, what are the risks and rewards of building a translation company this way?
It will not surprise you to learn that human beings are more challenging to manage than numbers and files. We learned early on that there was an optimal balance of in-house staff and language professionals. As time went on, the ratio of translators to project managers kept increasing. That was a function of more efficient operations as well as our rigorous training and certification programs. Hiring thousands of freelance language experts does not mean that interpersonal relations with our linguists are more distant, even if they live all around the world. We’ve integrated management tools and best practices to knit human relations together. Freelancers are essentially no different than employees in many ways. It’s just the contractual relationship that differs.
Sure, it’s a relief for a translation company like Tomedes to have a smaller burden of taxes, insurance, and paying “social benefits” by hiring mostly independent contractors. We accept that with lower financial burden comes less “forced” control over our workforce. But we’ve learned to compensate for uncertainty by training and certification processes as well as by the way we manage project teams. It’s also essential for our professionals to feel they are being treated with respect and understanding. Once the expert network gets big enough, there is more flexibility in assembling virtual teams that perfectly match the skills that a specific project requires.
Has your translation company adopted AI software in addition to human expertise?
Yes, Artificial Intelligence has revolutionized that translation industry. Linguistic neural networks and natural language processing, introduced on a large scale in the last decade, have rewritten the rules for how translation is best executed. Google Translate, for example, now supports more than a hundred languages, not just for text translations but also for more exotic tasks like camera translation and simultaneous voice interpretation. While the best human translators still can beat machines for more complex and creative texts, the machines can beat mediocre translators, and can do a better job with many high structured texts.
So Tomedes has adapted to the changing situation by integrating Artificial Intelligence and machine translation when appropriate and advantageous, while maintaining our core assets in our relationships with expert linguists and our ability to create top performing teams on demand, in response to the exacting requirements of our thousands of clients worldwide. Machine translation is not the enemy, but neither is it a panacea. In building a translation company, like anything else, it’s a matter of balance and agility to adapt to changing conditions. That agile approach has enabled us to scale and thrive even in challenging times.