AI-Assisted Translation – AI Powered, Human Perfected

Welcome to the Tech Talks Daily podcast, where you can learn and be inspired by real world examples of how technology is transforming businesses and reshaping industries in a language everyone can understand. Here is your host Neil C. Hughes!  

Hello and welcome to this episode of the Translation Company Talk Podcast, we are back after a short break as I was very busy in September. September and October are generally busy months as people return to work from holidays.

Conferences are in full swing and work piles up quickly. 

In any case, we are back with another episode, and today I’m excited that we will be covering the subject of embracing. 

Automation in the language translation process. 

I’m sure you all know that AI has been around in the form of machine translation in this industry for what it feels like a couple of decades already, we will be exploring automation beyond that, in our conversation today. 

My guest today is Matija Kovač.

He has successfully founded two companies, a language school which became to be the biggest language school in Slovenia and a translation platform. 

His first love is languages. He has studied Chinese language and spent some of his time studying in China. His other love is development. He’s been developing since high school and is the development father of Taia. 

Matija and Marko founded Taia to enable a more secure and time efficient translation process, seeing that sending files over email can be a security threat. They decided to create an app with a huge emphasis on keeping the files secure.

Since its first launch, the Taia app is consistently developing, creating new features to enable companies and an easy ordering process,  while being able to oversee all projects, the translation process is supported by the latest translation technology, which enables getting translations up to three times faster than usual in a cost-effective manner. 

Welcome to the Translation Company Talk, Matija. 

How are you today? 

Hi, my name is Matija, and thanks for having me on this show. I’m the co-founder of Taia Translations. I’m also the CTO and I’ve been an entrepreneur for about eight years now, something like that. I’m a father of two and a very avid coder and nerd whenever I’m not running a business. So, that’s it.

Feels like there’s a bit to unpack there. A coder, a nerd, and if I take you back before Taia Translations, what’s your origin story? Where did your passion for tech come from and what was it that put you on this path you’re on?

I guess I’ve always been a bit of a nerd and a coder, but I deviated away from this path in high school for quite a while, and then I studied languages. I lived in China for a while and I worked a lot in marketing and media, so I learned quite a lot of things from many different fields. I also grew up in a family of entrepreneurs with graphics in focus, printing and everything. This was something I did a lot of when I was younger, and then at some stage I wasn’t really happy with where my career was going, so I decided to call a few people, ask around who might be interesting to work with, and one of them was my current co-founder, Marko, who already had a successful company back then, and he proposed that we build a language school. 

So, we started with that since I’m good with languages and I know how to teach, I’d been doing a lot of teaching back then. And Marco already had a sailing school out there, so we started a language school. It quickly kicked off, became one of the biggest in the region, and we started onboarding a lot of bigger B2B clients like banks, insurance companies and so on. 

So, some of these companies came back to us saying, ‘Hey, we love your team, we love how you’re doing your language courses. Would you be willing to do our translation services as well? So, we said, let’s give it a try, and we started working in the translation industry as well. And that’s how we started Taia Translations in 2018.

Wow, what a great story. And of course, it was that path that led you to Taia, which is the translation platform. But what put you on my radar is you’re helping businesses translate their content more efficiently with the help of artificial intelligence and experienced translators. So, I love how you’ve fused the two together there, but could you just introduce the listeners to Taia, and what makes you different from, let’s say, other translation platforms?

Yeah. I think you nailed it quite well here. What we try to do is we try to help companies grow as fast as possible into new markets and getting their content across into other languages as effectively as possible.

When we started out, originally, this platform was only meant for helping clients to order and monitor their translation projects more easily, and more efficiently. But it ended up being a full-scale platform where you can either order fully automated translations of over 70 different file types, or you can use their tools to translate by yourself using AI assistance and translation memories and everything, or even outsource it to freelancers that you already have onboarded. But if you don’t want to deal with it and you just want to have someone else handle it, just drag and drop your files, select what kind of budgets you can fit it in and what delivery times you need, and our team is already working on it. 

It’s a very streamlined, very automated process that makes it much faster for us or for any of our clients to get their content across into another language.

And what’s the role of artificial intelligence in all that? Can you expand on how that comes into effect here?

Oh, yeah, sure. Well, the main technology in the background here is called NMT or neural machine translation. It’s basically AI networks that help predict what the translation should be, similar to like in Google Translate or DeepL, for example. Except that here it’s packaged with a lot more around it, helping it to perform better.

So, for example, the system is able to learn from your previous translations and propose existing solutions that you already have in your database, or even improve the translation AI model to produce better results within your existing one. We also do a lot of file handling. As I mentioned, around 70 different file types can be automatically extracted, translated and re-injected back into the original format, so you don’t need to do anything in order to have a fully designed document translated back into its original shape.

And there has been a lot of talk recently about AI and automation replacing jobs, but I often say that it’s more about how the two can complement each other, and that’s where the magic happens. But how are you seeing AI and human translators complementing each other?

A few years ago even, the translator had to do a lot of manual work. They would be presented with a source text, and they would have to manually type out the target text. What we do now is we give them a recommendation on how something should be translated, and they can either go about and change it entirely or just confirm this segment. So, this can, of course, lead to some issues, but what it generally does is it speeds up the process dramatically. 

This is just one point, but what it means in terms of how translators work is that now they’re able to work much faster. They can go through documents much more easily, and they can actually dedicate more time into making sure that the translation is correct. So they also get a lot of recommendations from the system – which phrases to use that have already been used in the past, for example (we call this the glossary), or using the exact same translations as we had in the past to make sure that things are consistent across multiple projects, in multiple documents and so on. When it comes to how machine translation is taking translators’ jobs; it’s actually not that black and white. What’s happening here in this industry is that the industry is growing very fast and it’s growing even faster due to machine translation. 

Because of all the AI assistance, so much more content can now be translated that so many more companies are looking into getting their content translated. The demand for translators is actually rising. So it’s an interesting trend. Let’s say you run an e-commerce platform, an e-store, and you want to get your content translated. About 10 years ago, this would have been a huge impact on your costs, right? But what you can do now is you can only have certain products and certain pages translated by humans, whereas the rest of your content can be translated solely by AI, and it’s much, much cheaper. The quality might not be perfect, but you don’t need 100 percent perfect quality for all of the products in your store. You only need it for the top 100, maybe 200 selling products. 

So what this means is that more and more companies are looking into getting into other markets and getting into localization because it’s much more available. But this also drives up the demand for human translators, who perfect some of the content on the way to the customer.

And as someone that’s working right in the heart of the industry, how are you seeing AI effecting the localization industry and what do you think we can expect in the future? Are there any trends?

Yeah, the main trend we’re seeing is that a lot of companies are trying to adopt this technology, and the ones who haven’t or are refusing to will most likely be swept away by this wave of automation that’s coming. So we see a lot of different business models. We see a lot of different technologies emerging. Even since we started, the industry has changed tremendously. 

But still, most of the current translation needs are still served by traditional translation agencies, LSPs, or freelancers, where they usually don’t rely that heavily on technology. So this means that the companies out there who plan on going into localization and find the good partner who’s going to be able to support them with a lot of automation and technological support – which obviously brings down the time and the costs required to get into a new market – will have that competitive edge compared to the companies who don’t pick their translation partner as cleverly. 

So if you plan on getting into other markets, make sure to find a localization partner who’s capable of supporting you and won’t give you even more work. So, for example, a very blatant sample that we usually see is companies copy-pasting their websites and content into word files and sending them via email and trying to manage all of these projects manually and trying to wrap their heads around all the billing details and whatnot. Whereas with a platform like Taia, everything is more or less streamlined. You can integrate the API or just drag and drop your files if you already have them. And the platform takes care of most of the processes for us and for the customers, so everyone can benefit from doing less repetitive tasks and getting more meaningful work in.

And to counteract all those negative news stories we’re seeing on our news feeds around AI, etc. in the future, what excites you about the future of translation and the role that technology will play in that too?

It’s the stuff that we’ve seen in the last few years. It’s tremendously exciting to me. You’ve probably seen GPT-3 making the rounds as one of the biggest models ever built so far and being able to not only translate but also produce text and change text and do whatever you can imagine with a certain text. 

So, I think these huge models and these new advancements in AI are about to bring in stuff we don’t even expect yet. You know, we will be able to achieve a very high level of AI, nearing general intelligence actually, in a matter of decades, if not even sooner. So I’m really excited to see what the future brings. The only problem is that the further we go, the more complicated these things get. So as we are, you know, climbing up this development tree or discovery tree, however you want to call it, it’s getting harder and harder for people to get into the industry and build something new because you need a lot of resources, not only on manpower and people with the right skills and the right knowledge, but also computational power. 

Models, like GPT-3, can be only powered by about two or three companies in the entire planet that can support such infrastructure, that can pay for all the computers that you need to power it. What’s happening is that companies with a lot of resources are able to build really high-end AIs and everyone else will probably be left behind with not so capable AIs as those who are in power and have the resources. So I think it’s going to be a weird time coming for us.

Fantastic. And another question I wanted to ask on behalf of startup founders that will be listening all over the world – maybe they’re much earlier in their journey than you are right now – what kind of team do you think you’d need to build a successful startup right now and what is most important when building your corporate culture? Is there anything you can share around that from your experience?

Oh yeah, I’d be happy to. If you’re building a tech company, obviously you’re going to need some tech folks first. And this, I think, is a really crucial factor that you need to take into consideration. So if you’re not a coder yourself, finding a co-founder, a partner, that can at least work around with code and talk to people who are coders and understand everything that’s going on, it’s going to be much, much easier for you to get the company going. 

The one thing that we forget a lot of times is that building a good product, and finding a product with a good market fit as well, will take a lot of your efforts from marketing and sales off. If you have a good product, it’s going to be much easier to grow it and to scale it than if you don’t. So try to think ahead and build stuff already when you’re building your MVP, which will be scalable and that will not lock you into a certain tech stack that’s not scalable or that you cannot later on evolve, because starting from scratch is very expensive. So even when you’re building your MVP, keep in mind you’re going to have to grow this MVP into a beta version and then into a 1.0 production version and so on. Yeah, that’s definitely important. 

When it comes to company culture, I’d say we’re getting quite a name for it ourselves. We’ve grown from 10 or 11 people company to almost a 30 people company in under a year. So there’s been a lot going on at Taia in the recent year and a half. And I really love this team. I don’t think I’ve ever been working in a team that’s as cool as this team is, and I can really rely on basically everyone in my company to do their best and make sure that we’re going into the right direction. 

What’s important for us, for example, is a lot of autonomy. We like our people to think on their feet and not be told what to do, but rather them telling us how they’re going to solve something, how they’re going to do it. So finding people for your company who find problems and know how to solve these problems, and who can strategize and lead with example is really important. So if you’re building an amazing startup, surround yourself with amazing people. Don’t go with the boring and uninteresting.

Neil C. Hughes:

Fantastic advice. And what about yourselves? What’s next for Taia Translations? Is there anything you can share about what you’re going to be focusing on?

Matija Kovač:

Right now we’re focusing a lot of our energy into improving what our platform can do. So we’re increasing a lot of our SaaS capabilities to be able to expand more as a SaaS company. We’re doing a lot of experiments in marketing, so trying to pinpoint what kind of messaging works best for which target audience. We do a lot of – I call it science. 

So basically, because we come up with a hypothesis, we shape and experiment around this hypothesis, we go and run it for a few days or weeks, depends on the experiment, measure the results and with some data driven decision making then go into another round and another round and so on. So we keep on experimenting and improving this way. And the major thing we’re focused on right now is also getting the next round of funding to be able to support the fast growth that we did in the last year or two. So we’re currently out in the market for a lead investor. 

We already have some pledges for our next round. We call them the seed round. But, yeah, we’re raising about a three million euro round this time, and it’s going to help us to perfect the platform even more, but mainly to invest in sales and marketing, acquire a larger chunk of the market and see where the traction brings us.

Neil C. Hughes:

Well, I wish you the best of luck in 2022 and beyond, but before I let you go, we started talking about your origin story, what put you on this path and the startup journey you’ve been on. I’m now going to ask you what has been the soundtrack to all that? Is there a song or piece of music that has inspired you, helped you in your career or just helped you get your head in the zone before going on stage? What would your song be and why? We’ll add it to our Spotify.

Matija Kovač:

I think my Spotify account is definitely convinced that I’m a lunatic because I have everything from Iron Maiden to Spanish folk music to minimal electro techno stuff. It’s a weird mix, so yeah, it’s hard to put a finger on one single song. Wow, you got me there. I didn’t prepare an answer for this one.

Neil C. Hughes:

I spoke to someone the other day and they were into pirate metal. I don’t know if you’ve heard that?

Matija Kovač:

Pirate metal? Sounds good. That’s good, yeah. I have everything here, yeah. I don’t even want to start reading from my Spotify. It’s just all over the place.

Neil C. Hughes:

What about when you coding? Are you an instrumental dance kind of guy, or?

Matija Kovač:

Yeah. Usually when I need to zone out, it will be like some kind of chill step or some house music or something like that. I like this very cosy electronic music for zoning out. But if I need to plough through a bunch of tasks, I’m probably going to go with some heavy metal or even progressive metal or stuff like that, even some older dream theatre albums or stuff like that.

Neil C. Hughes:

Well, now we’re talking. Before I let you go, for anyone listening and wanting to find out more information about anything we talked about today, find you online and contact your team.

Matija Kovač:

Our website is out there, so it’s Taia.io. Just Google Taia Translations and you’ll find us. And we’re always happy to chat and to help when it comes to localization. And you can find me on LinkedIn. I get a bunch of connection requests every day, but if you’re persistent, I’ll definitely find you and get back to you.

Neil C. Hughes:

Awesome, love so much about what we chatted about today, from the future of machine translation, how localization can enhance global strategy, how AI is revamping the translation industry, and also for proving that very important point that AI and human translators are actually complementing each other and not competing with each other. And I think that’s such an important message, too. And more than anything, thank you for sharing your story with me today. So thanks for sitting down.

Matija Kovač:

Thanks for having me, Neil. It was a pleasure.

Neil C. Hughes:

What a great guy. And I do quite fancy a trip to Slovenia, too, so I’m hopeful to meet him in person in the not-too-distant future. But a big thank you to him for coming on and sharing his entrepreneurial story there and insights. 

And as always, as every podcaster does this, I don’t do enough as I should because I don’t want to annoy you all, but if you can rate and review this podcast and help us in the fight against algorithms and bring more listeners to our community and prove that technology works best when it brings people together, please do a quick rating and review wherever you listen to the show. 

Send me a screenshot once it’s out there and I’ll give you a shout out and a much deserved thank you from me too. Meanwhile, I’m going to go off and prepare another guest for tomorrow’s episode. So a big thank you for listening, and until next time. Don’t be a stranger.

Outro:

Thank you for listening to the Tech Talk Daily podcast with Neil C. Hughes. Remember, technology works best when it brings people together.