S03E10: Automation Across LSP Business Functions

John Moran

Thank you, Sultan. It’s it’s an honor to be asked.

 

Sultan Ghaznawi

Appreciate it, John give me some background and introduce yourself to people listening about yourself and their company.

 

John Moran

Sure, sure. So I guess my background is both on the IT side, so mainly software development and also on the linguistic side, so I have made a living in the past as a German to English translator and I don’t translate a lot these days unless they’re, you know, short marketing texts. But I do spend about 1/3 of my time reviewing German to English, a lot of the time, you know, reviewing texts that were originally machine translated and then post edited. And so I have to make sure that the work that we send to our customers is acceptable.

 

So, yeah, so I mean that that’s pretty much and I run a translation agency. So we could be characterized as a specialized language vendor and which is kind of an industry insider way of saying that many of our customers are MLVs or multi language vendors in other words much larger agencies that have you know teams of salespeople and we’re a small company, we have a staff of five plus plus our freelancers so our staff are mostly project managers and yeah, we specialize in German to English, so we do German to English, usually high volume projects if we can. It’s, you know, it’s how we add value so we’ve had projects where we’ve had, you know, between 30 and 70 translators working in parallel. And which is an interesting war story, you know, going from from the lower number to the higher number.

 

And then we also do Irish and we can talk a bit about how I got into Irish uh, a few years ago. It’s a different market. We rarely have more than three or four translators working in parallel, but we always have translators working for Irish. So it’s a, it’s a language that I’m, I’m Irish so I have, you know, emotional feelings about, but also, yeah, it’s small, small side of the business. It’s not what we do mainly, mostly it’s German, English and or other language pairs. We can talk a bit about that, but yeah, so that that just gives you a rough idea. And so and also I should say that our Irish work is a mixture of customers and direct customers. So most of our direct customer work tends to be either Irish or just sort of sporadic languages that that our direct customers send us.

 

Sultan Ghaznawi

Quite interesting. How did you decide to enter this interest in this industry? I know you have a story to share. Was it by choice or accident? How did that happen?

 

John Moran

Yeah, I I don’t know if you’ve read Renato Beninato’s book, The General Theory of Translation Agencies, I think that’s…

 

Sultan Ghaznawi

The Translation Company…

 

John Moran

Yeah, the translation company, it’s yellow. I have it out there somewhere. And in it Renato comments that nearly everyone he meets fell into the translation industry by accident. I, I know that, you know, I’ve somebody asked me that question before and I had to think about it. I think the first time that I that I fell into it, so the thing that really triggered it was I was uhm… I graduated, and one of I did quite well. So I graduated top of my year in German, which was a minor subject with my degree in computer science. But nonetheless, we did the same exams as the Germanic studies students who were just studying German and other Germanic languages. So my German was pretty decent when I left college and I was walking up the stairs of the arts block in Trinity College, Dublin where I had studied. I was a post grad student at that point and one of my lecturers, Dr Gilbert Carr, stopped me on the stairs and said would you like to teach the class in technical translation and that you took, that I had taken with him the year before because I I had really enjoyed the class and I think he saw that, so I ended up teaching this class in technical translation and also it was technical and semi literary translation was the title. And then, so I was ironically, I think a lot of people end up lecturing in translation after working as a translator for years.

 

For me, it was the other way around this was my first sort of paid gig and it was just a part time lecturing position, but on the back of that a company called Draytek, who no longer exist, a subsidiary of Cog… of Commerzbank, which is a German bank who also know sorry of Fresno bank which is now Commerzbank, and they contacted the Germanic Studies Department in Trinity and asked them if they knew anybody that knew how to translate it it documentation from German to English they put me in touch with them and I didn’t know how much to charge per word. I had no idea that there was such a thing as the translation industry. So I took an exam that I had taken and the amount of time that we had for the exam. We had an hour to translate something like 200 words. I based it on €500 a day because I’d been paid that much to do a programming training course the same week and I ended up with a word price of about $0.40 per word and that’s what I charged for a couple of years to this customer who… so it was a pretty decent, I think we can all agree that’s that it would be great if we could get you know those prices normally these days. But the the outcome of that was because I was charging pretty, pretty decent rates per word I was able to find a reviewer to work with in Dublin I could easily afford him. And so I did the translations myself at the start and the customer was happy and but I wasn’t really happy with it, I felt I was, you know, making mistakes and then I I put the reviewer in the process and I really learned a lot from the feedback I got from him.

 

And then all of a sudden the customer needed a lot more work than I could possibly do. I I had other commitments, I was teaching, I was working on a PhD. So I hired a few translators and I set up a website transpiral.com. And yeah, I mean the rest is sort of history. I did that, I we, only had that one customer dry tech for about two years and then language came on board and I grew with Lionbridge, so and one of their project managers, Ingo Shuman, who’s no longer with Lionbridge and told me that for a number of years and that that we were working together, we were Transpiral, was by far Lionbridge’s the main supplier for German English as an agency. Otherwise they worked with a lot of translators. And that’s really part of the business model, as a specialized language vendor, your customers are always working directly with translators, but they might work with one or two agency vendors for bigger projects, and that’s where we like to fit in.

 

Sultan Ghaznawi

It’s very interesting and I’m sure you have a lot of stories to tell. Can you share some of your interesting stories about your experience in this industry? What were the things that stood out to you?

 

John Moran

Well, so I think one of the things that’s a little bit unusual about me is that or about transparent is that, uh, so what happened was that Lionbridge and Bowne, one of our other customers merged and they closed down all of the offices in Germany. This is about 19, sorry by 2000 and I think 3 is that right or 2006 anyway, so it was in the mid naughties and that left me with, you know, language had been pretty much 90% of our sales and I decided OK, well, you know what I the computer science degree I used to work as a programmer. I’m just going to go back and work as as… so I did that and I I worked as a programmer for a number of years, mostly for for a subsidiary of Siemens, Nokia Siemens Networks. And and then I got a little bit bored of programming and I thought, you know what would be really cool would be to combine what I’ve learned as a programmer with translations.

 

So I decided to go back to college, to go back to Trinity College and I was a PhD student this time in computer science, and I did uh, some research which has been published. I can send you the links after if you want if you have show-notes… it’s an on the topic of measuring translator productivity and So what I did was I downloaded the source code for an open an open source CAD tool called OmegaT which is programmed in Java which is my preferred language programming language and I added logging statements to the tool and then I also programmed a replay function and so you could replay what was happening in the CAT tool, so let’s say for example the translator was typing the target sentence and decided to hit the delete key 5 or 6 times that would show up in the in the in the replayer it would show. In other words, it would show up in the logging, but in a normal CAT tool, you wouldn’t know that the translator had been hitting the delete key, you just see the target sentence. If that makes sense.

 

Yeah, so it was basically kind of a trick to, you know what we called an instrumented CAT tool. And I demo’d that to an outstanding engineer called Dave Clark who works for Welocalize, and I remember Dave saying in the meeting that that fills the gap in our thinking in terms of post-editing pricing, because post editing pricing really boils down to the question, how much faster is the translator with machine translation, right, as opposed to translating from scratch? So we ended up doing, I think if I’m not wrong, and somebody I’m, I’d be happy if someone would tell me if I’m wrong. We did this work in 2013 and I think it’s still the largest or certainly one of the largest published productivity tests. And you know, at, you know, in, let’s say in the public domain, obviously, you know, companies do all sorts of internal research, but they don’t necessarily publish and at conferences or, you know, online. So yeah that was that was a really interesting year or two, I got to work with some really stellar people, Olga Bergavaya, who’s well known at conferences, I saw her picture on the front page of Multilingual Magazine recently and then Laura Casanellas as well, their MT program manager.

 

So I got to work inside of a company that was not unlike Lionbridge, let’s say. Let’s say global MLV, many of whom, many of whom you know many of the staff had worked at Lionbridge as well. So I got to, I got to see the inside of one of my former clients and and I have to say it’s I remember driving to work just you know singing in the car. I used to really enjoy, yep, and and I learned a lot about, you know, machine translation and how to manage projects with post editors and and I mean, we can talk about that a little bit further on in the podcast and because I think it isn’t, I think it is a question that that I mean it’s it’s a question that I still ask myself as a as an agency owner these days you know, why is the industry so far behind in terms of supporting post-editing in the tools when it could actually, you know it’s a problem that sort of should have been solved in in my opinion in 2013.

 

Sultan Ghaznawi

Absolutely and we… I think we will touch upon that in one of the questions in our interview today. But let me get an overall perspective of yours, let’s dive into our core topic… that is, we cover topics that affect language industry providers, I think there is nobody who hasn’t heard or been involved in automation in our industry in the past decade or so. So, give me your personal overview of what automation and a language service provider looks like today or should look like. I mean, you can’t cover all the angles, but whatever you can think of, please.

 

John Moran

OK. OK. So I mean let’s imagine the base case where you’ve got no automation, I mean go back to 2000…when did I start out? In 1999 or something, right… and you know when the Prince song came out. That was Trados 2.0 if I remember correctly. I remember as a student I got an e-mail address and no one knew, no one had heard the phrase e-mail before, and certainly no adult had, so I thought it was really cool that I had an e-mail address that my parents didn’t. And, you know, so I’ve seen the industry develop over the last, let’s say quarter of a century and I think that if you so, so, so in a sense that’s the base case, right, basically. You know, at some point, I mean, riding a horse is… a technology, right. So, that’s my own personal sort of base case and how has it developed over the years? I think we still do largely rely on e-mail. I mean with the caveat that almost all customers have a portal of some kind and it’s either been bought or it’s been built. In other words, some customers have their own bespoke portals, and some customers use, I don’t know, Plunet or XTRF or whatever they use.

 

So we’ve seen certainly automation in the sense that, you know, portals have become, I don’t know if I could say more sophisticated, they’ve become more of the norm, right. I mean every, every single one of our agency customers portals frustrates me in one way or another right.. I, you know, it’s very hard for me as a programmer to look at a piece of software and think, you know, it’s perfect and but I’ve certainly seen portals I think to myself, yeah, this is actually really saving us a lot of time. So I’ll give you an example as as an SLV, it’s great for us to be able to just go to a portal, click on a button and it does the invoicing for us. And you know Ideally, if we can download some sort of CSV file so that we can mirror the information in our own system to make sure that we’re paid properly, that’s perfect too. But certainly I think, I think automation in terms of billing has has been is a trend over the last few years and it’s one that we’ve tried to mirror within Transpiral.

 

We have our own system that I I didn’t program it myself, but I I designed it to create, you know, the work orders for the translators, and we have our own sort of I would say a fairly basic sort of sales process. And so that’s one kind of automation and then the other type of automation and particularly the one that’s interested me as a researcher or as you know as a writer, let’s say in terms of the papers is you know, how do you translate more words per hour without affecting quality negatively. So there are really 3 answers to that question. The 1st and most interesting one for me, not so much today but maybe a few years ago, was and the use of speech recognition software. So we we work with a number of translators who were using Dragon Naturally Speaking, and I was really interested in that. And I I’ve given a few talks at conferences on particularly translated conferences on that topic, and machine translation, of course, is fascinating. We’ve seen and really big jumps in terms of machine translation quality in the last five years, I would say. DeepL stands out as a system that even I was at a uh, met up with some people privately 2 nights ago. And you know, we were talking about DeepL and so it’s pretty well known system even outside of the industry.

 

But and then there’s the third option, which is predictive typing and there are some, some companies that have done some interesting work around that. There are lots of different types of predictive typing. You’ve got sentence completion, predictive typing. You’ve got sort of clumps of words. You know where it just sort of randomly chooses to give you a few words as you’re typing. And you know that that that those are functions that are largely supported by the CAT tools. Sometimes it’s just the predictive typing is just it finishes The word for you from The from the term base. So these are All these are all technologies that I was very interested in and I don’t do it anymore, but certainly I spent some period of my life full time measuring the impact that these technologies had on professional translators, including myself.

 

Sultan Ghaznawi

So, that’s interesting and there’s this notion, John, that anything that can be repeated can also be automated, whatever those things inside a language service provider. And is that even true? Can that happen?

 

John Moran

I think it depends. So, you know, it really depends on the budget. If we’ve got, you know, if we’ve got a large project with lots of words coming through per week and then obviously we’re more inclined to want to spend programmer time, which is we have 3 programmers in the business, including myself. But I prefer not to program. I find it very difficult to think in terms of being a programmer and managing a business. But we have, you know, we have programmers in in India that that… two programmers we work with in India. We have our production manager used to be a programmer, and I also used to be a programmer, so we do tend to look at, you know, how to solve problems from a technical perspective. But we do have to ask ourselves the question, is it really worthwhile doing this? I mean, you have to ask yourself so with any kind of project, is it worth taking on the project at all or once you’ve done it for a few months you realize you’re not making enough money. Maybe you need to tell the customer that you know they need to find another solution and I think that’s where we do see an automation in terms of trends. I would say we’re getting a lot of value at the moment from a system called BeLazy, and I don’t know if you’ve heard of BeLazy. It’s Istvan Lengyel’s baby, I suppose. He started the company, he was Istvan, was I’m sure you know the CEO of Kilgray, who are now MemoQ and so what BeLazy does is it helps you sort of drill down. Well actually what it really does is it connects to various, let’s call them ERP systems, sort of management systems within large agencies or some of the ERP systems that you can buy like Plunet or XTRF… And I think he was at the Plunet conference yesterday.

 

And so that helps us manage projects where we don’t have minimum charges. So for us minimum charges are a big a big factor because we’re in SLV, we’re working with MLVs and of course they’re selling project, you know, they’re selling large accounts to large customers, which sometimes come with the requirement from the customer not to charge minimum prices for smaller projects and so where you where you don’t have minimum price if you’ve got, I don’t know, 10 words charged at 10 moon dollar cents, so moon dollar cents isictional currency, yeah, a lunar currency, unless anyone accuse me of… but it’s a good it’s a good mechanism for talking about, you know, economics. So, so if you’re, if you’re selling, you know, 10 words at 10 moon dollar cents, then you’ve got one moon dollar. And if you have to create a purchase order for us, for what we call a work order for a translator and you have to issue an invoice to a customer, you’re going to spend 15 minutes doing that. So you’re you’re earning, you know, no money, you’re losing money, so you have to automate. If you don’t automate, then there’s no point in taking on on the account and I think we’re seeing more of that happening and the automation of small hand offs with large hand offs we had a 20,000 word hand off 10 minutes before this call. I’m more than happy to send an e-mail.

 

Sultan Ghaznawi

So the OK, I see your point. You want to automate things where you know there’s no value in assigning an actual human to process it when it can automatically be processed by computer processor.

 

John Moran

Correct, though of course the only human touchpoint is the translator. Now I will say that that’s the ideal situation. But the reality is that, you know, at least a couple of times a week we get emails from a project manager in South America to say, you know, these five sentences are three hours over due and because we have same day deliveries and you know, we we manage that as best we can and so, so it’s not fair, it’s not true to say that in the case of that particular account that the that everything is automated. We do have some accounts where I honestly don’t even think about it. You know from one month to the next, because everything is smooth and has been smooth for years. And I’d love to have more of those.

 

Sultan Ghaznawi

John, let’s go through the processes and drill a little bit deeper. The core of an LCD production system, how many of those processes have already been automated and you just mentioned BeLazy and Istvan’s project, how many or which ones remain unexplored and should be automated?

 

John Moran

Right. Yeah, I mean I had a very interesting conversation a few years ago with Henry Dotterer from Proz. I posted a message on the Proz board trying to elicit some idea, some feedback from translators and unfortunately I must have caught somebody on a wrong day because I just got the response so with all these people trying to take our jobs or, you know… it was a lady that I then realized had a history of, you know, being a bit histrionic. But anyway, I posted this this, it was just a one sentence question to say, hey translators, would you guys be interested in this kind of software and Henry contact would be on the back of it, but I do feel I mean, I never got around to implementing it, but I do feel that a lot of time is wasted managing capacity. And you know if we look, if I look at across our MLV customers, we get emails all the time telling us you know are you taking Easter holidays? Are you taking Christmas holidays? The answer is no and no, because we have both Christian and Muslim and Orthodox Christian people working in the business and we’re quite careful to make sure that we have 365, I won’t say 24 hour, but we certainly have an 18 hour coverage. Uhm, so, you know, I think that agency spends more time than necessary managing capacity. I think if translators were, for example, if they were to use the calendar function on pros.com more actively, or everyone decided to use Google Calendar and I I just feel that that’s a that’s an area that PM spend too much time on.

 

Now, it’s not that simple because I mean if we have a project with let’s say 70 translators working in parallel, then yes, we’re really interested in knowing what their capacity is on a day-to-day basis. But when that project is finished, we don’t care what their capacity is, right? So uhm you know, but I mean that is the answer to your question I think, I think, I mean we’re going to see increases in the quality of machine translation certainly even moving into other languages I’m sure Arabic, I know Arabic is you know machine translation quality is it’s pretty low compared to let’s say Italian from English. So we’ll see, we will see improvements there but they’re kind of, they’re sort of obvious and based on you know past performance and I would say in terms of workflow, I think capacity management is definitely the area where PM’s are spending a lot of unnecessary time, and that could definitely be improved somehow. And if anyone wants to talk about that, they’re more than welcome to reach out to me, actually I have a budget in mind for a project that that might help that.

 

Sultan Ghaznawi

So Speaking of machine translation, you just brought up automation in the form of MT or the speech to text has been around and into a good degree successful. How is the industry embracing this new area of productivity?

 

John Moran

Well, in terms of the language services industry and speech recognition absolutely zero. I can honestly say that and many translators are using dictation software, and that if you talk to agency owners, they’re barely aware of this fact. And now you have to be very careful with speech recognition because as I said, I spend maybe, you know, quarter or third of my time reviewing the work of German English translators. These days most of it is post editing, but in the past we worked with a number of translators who were using dictation software, and I used to give people training sessions on how to do that because I wanted to see them earning more money even if it made no difference to my by business. But you do have to be careful. I’ll give you an example.

 

We worked for a company, German agency. We sent the translation back, and much to my disgrace I missed one of the errors and it was the word chemical was mistranscribed I can’t remember it was some sort of medical term and like pregnancy or something like that, but it was a word that sounded more like chemical and it totally changed the meaning of the text in a way that generally speaking, mistakes in machine translation post editing don’t. I mean most mistakes in machine translation post-editing are such that the style of the target text is a bit wooden, and sometimes, you do have to be careful because you’ll sometimes see semantic reversal in the sense that the word no might be left out by the machine translation system. But in general you won’t see these kind of catastrophic dicto what I call dicto-errors as opposed to typos. So so that’s that’s the speech side of things.

 

And in terms of machine translation, I mean I think machine translation has it, I think Translated in Italy, had a website a few years ago called the Big Wave. And it really is a big wave. And certainly in my own business, I see post-editing requests for maybe 40% of the work that comes in you know when it might have been. 5 or 10% three years ago. We generally, without MLV or with our agency customers, we try to be as transparent as possible, so we won’t use machine translation for it or not or unless we’ve gotten you know written permission to do so because you’ve got the problem of potential IP spillage uUnless you’re using an internal system, and we don’t actually have an internal system because quite frankly it’s very difficult to build an internal system that’s as good as DeepL.

 

So yeah, I mean, we’re seeing uh, quite an increase in post editing and we’ve done you know, I mentioned actually the the project that we did that had 70 translators was actually for your last guest, it was for ULG, I don’t think they’ll mind me saying that. So I think, sorry that sorry, excuse me, that project was 70 translators, but it was translation from scratch because it was for a court case and we didn’t want to take any risk with IP. And since then, we’ve done a number of big projects in the automotive space using machine translation and I’ve been feeling that from now on, anytime we do a big project with more than 10 translators, it’s going to be post edited because frankly, it’s just more, it’s just more efficient. Uhm now in contrast to that, we also do Irish and I would say twice a week, well, not twice a week, maybe twice a month I send emails back to customers telling them that we simply don’t do post editing for Irish. Yes, google provides output for Irish. But it’s not good. The translators generally speaking, will either refuse to work with it or they won’t offer any kind of discounts, so you’re just taking a risk for no reason.

 

Sultan Ghaznawi

Yeah, John, can you give me some examples of case study or case studies for successful implementation of automation strategies and in the language services companies you deal with larger MLVs, but also you have other peers who are and within you know the sector. Can you tell me how both of these have implemented automation successfully to their benefit?

 

John Moran

Well, I mean going back to what I said before, I think that you know automated billing is really what I see most of in terms of the automation on the, on the part of our customers. I mean it I don’t want to be negative, but I will say that, you know, I think that really one of the things that large teams need to think about is OK, it’s fantastic if you’ve got a really configurable, really like space age management system, but it does nothing if you’ve got if you hire somebody who used to work in a bar 2 weeks ago and hasn’t been given any training in how to use the system, yeah. And so I think that I think that’s probably if the one thing that I’ve seen throw up problems is project managers who haven’t received the training in the system that they need to receive to use to use after well now that said, most of the project managers and work with her very good. But of course, we’ve had a few war stories over the years, and I know I left a note for myself. That’s one thing.

 

But then the other thing that kind of frustrates me and is that it’s very rare for any of our agency customers to actually ask us, you know, what do you think of our software? You know, how do you think it could be improved? I mean, we are the users of the software, right? So I mean, I had a I had a conversation with the CEO of probably the second largest agency in the world at a at a Slator conference in London a few years ago. And I offered my time as a, you know, as a as a software consultant to try to help them with the software that I knew was, you know, because I’ve used the software, I know what the problem is, basically the problem is that the logging is not getting shared with the with their with their customer support teams. So they’re asking you if you’re asking for a screenshot of a piece of software, you shouldn’t be, you should just be looking in the log file at the time that the project was done to see what the software is telling you, if you know about the error that was thrown, yeah. Great, great.

 

And so yeah, you asked me for successful. So the successful ones are, the successful ones are the ones where, you know, we get a handoff, we hand back smoothly, there’s no, you know, there’s no e-mail exchange and we can, you know, invoice using a click or two. I would say out of the twenty or so MLV customers we have that describes maybe three of the systems.

Sponsor

This podcast is made possible with sponsorship from Hybrid Lynx, a human in the loop provider of translation and data collection services for healthcare, education, legal and government sectors. Visit hybridlynx.com to learn more.

 

Sultan Ghaznawi

John, let’s let’s shift gears here a little bit and put our flashlight on end clients, let’s say if we are not dealing with MLVs. How do they perceive automation? I guess they wouldn’t care about production of the translation if it meets their quality requirements, but do they care about our mission in the business process like project management?

 

John Moran

No, inn my experience our direct customers care that we delivered the project on budget and on time. And you know, at the expected level of quality, what we do for that to happen doesn’t make that much difference to us or to them. And now we have, we have a few private sector direct customers and we love them and but most of our direct customers tend to be government departments. So I can tell you, I mean, I can tell you kind of an interesting story about how we got into Irish and so I was an and that’s relevant because you know it’ll explain to you how we have so many direct government customers. So we were…. Transpiral specialized in German English as I said for language. Then I closed the business and we opened it in 2013 as a limited company, sort of, let’s say more intentionally this time to grow as an agency. And the reason I reopened the business was I had met a lady at a conference who was talking to somebody while I was having my lunch at the conference and I could hear that she was really selling more or less the same as what I had been selling, in Transpiral which was technical German and English. And financial so I had a quick chat with her afterwards about a presentation that Dion Wiggins from Asia Online had given and we exchanged business cards. I went home, I connected with her on LinkedIn, and I completely forgot about the conversation. Six months later, I had my LinkedIn open and I noticed a message to say that unfortunately, the customer had gone into liquidation or that the agency. It was a company called E-teams onn the West Coast of Ireland, where I’m from and there was a number for a liquidator, a solicitor, a lawyer, not to confuse our American listeners. And the I without thinking and this this costs me a PhD, so I really do think back to this moment in time. I reached for my phone, rang the number and spoke to the liquidator and he told me what money they you know, what they were looking for, what they had to sell.It was an XTRF system with you know, data about customers and translators. There were translation memories, there were and term basis for customers And an e-mail address that came with it so we negotiated a price and I bought the data of this company that had been in liquidation for over six months at this point, and I didn’t really think too much about it because I hadn’t really thought about it in terms of selling translation services. I was more interested in the time at the time in the translation memory because I thought, well, I’ve just developed a state-of-the-art system to measure translation performance with machine translation. Wouldn’t it be cool to build an MT engine and test against that and what was really funny, I remember the conversation with the guy, you know, I’m from the West Coast of Ireland, so I I understand the dialect pretty well. And he said that, you know, he’s, he’s just a West of Ireland solicitor and last week he was dealing, he was running around chasing chickens and now he’s selling translation memory which is not a conversation I ever thought I’d have in my professional life. So yeah, so that’s how we got into Irish and I I had studied. I had I, there was a time in my life I was able to speak Irish pretty well because I spent a lot of time in the summers speaking it. But yeah, so yeah, that that’s Irish and Irish is its direct… our direct customers are mostly government customers and the problem with government, government customers is that legally they have to send out requests to multiple agencies and the pricing can get very competitive. That’s kind of the problem. There are, there are a couple of agencies that will just, you know, machine translation text with Google Translate you know, pay for an hour of translator time and send it to the customer in the assumption that they can’t speak Irish and we don’t do that.

 

Sultan Ghaznawi

So, Speaking of which, do clients see the advantage of automation in creating new and innovative solutions beyond the traditional model of translation and localization, for example, we see that web applications and mobile applications, you know they they offer new areas of how people communicate. For example, if you were to look at WhatsApp, it’s an application, but do we as an industry plug into these new and innovative solutions that are out there?

 

John Moran

Yeah, I mean I think that the successful agencies are doing that and but I, you know, large agencies, I think, I mean that’s a general rule the larger you are as an agency, the more services you provide, right? And we tend to be quite specialized we we essentially what we see ourselves as is a turnkey solution for large volume I would say German to English, we get new customers for Irish every week at the moment, because Irish this year was made into an official EU language, but I will put it this way. If I find a new customer for German to English, I’m happy, If I find a new customer for Irish, I’m not. If I find it new translator for Irish, I’m happy. It’s actually very hard to find translators for Irish so and when I say translators, I mean, you know, really good translators that can pass our QA. So I I think that we, I think that we have seen ourselves provide… I mean if you look at two years ago, I think we actually made more money from layout than we did from translation and yeah so and machine translation has a role to play there.

 

So what we what we’re really good at is you can send us 10,000 pages of PDF and we’ll turn that into a word file starting the next day, and we’ll be finished those 10,000 word pages within a week. and if you keep sending us PDF files and we can scale up the project, then we can do about 100,000 pages and I think that was a record in a week. And you know, we’re working with partners to do that, but it’s partner that, you know, we work with every day. And what the effect that had in particular on the one project, one automotive project I’m thinking about was that the… we actually ended up sending these word files back to the customer who then created a Trados package. And because there was leverage within the Trados package, fuzzy matches 100% matches repetitions, internal repetitions and the layout cost was actually not just zero, they were making money by paying us to do layout, right? Because the alternative is to take a PDF file, and I mean you can try and ask the translator to create a table and or a Visio diagram and they’ll just say no.

 

And I mean, you know, we made that mistake in the early days, and so yeah, so I, I was I would say that you know, we’ve started to provide other services as well. We’ve done a little bit of transcription. It’s not our main, you know, our main thing. We’ve done some AI annotation, mostly for machine translation projects. We’ve done translation for machine translation projects. Actually, we’ve done a lot of that, because a lot of the conferences I go where I go to our machine translation conferences, you know, based on my degree being computational linguistics. So I mean, we are seeing more and more work outside of traditional, let’s call it TEP. I’m sure we will see more and I think that any agency would be well advised to keep an eye out for interesting services that can be provided instead of TEP, because quite frankly, we’ve seen the value of TEP go down in terms of how much money we make per word as a result of machine translation getting better.

 

Sultan Ghaznawi

So how about the translators and other stakeholders that produce the actual translation output? How does automation affect them? How are they responding to the automation movement?

 

John Moran

Well, it depends what you call automation. So I’ve never met a translator that didn’t like to go to a customer portal, press a button and the, you know, the invoice is created on 25 jobs. Yeah. So then so that kind of automation is always popular and machine translation I have had the honor of being asked to present at a number of machine… at a number of translator conferences over the last few years up to the COVID lock down. And I usually ask translators, do you prefer to work off in offline tools or online tools and I’ve only ever seen one translator who put his hand up and said I prefer to work in an online tool. And I think over the last few years I’ve seen translators sort of adapt to that and but generally speaking they prefer to work in there, you know, in MemoQ or in Trados or in OmegaT or whatever the offline tool is that they prefer. And online tools can…. you know the problem with online tools is that if it’s good, it’s good.

 

But if you are a translator and you’ve got 10 customers and one of them has a bad online tool, it’s hard for the translator to remember which tool is the bad online too, or which customer is the bad online tool. Then it might not be the customer’s fault, it could just be that their own Internet went on the blink for a few minutes and frustrated them. Yeah, so this idea of always on. But you know, it’s a trend. It’s definitely we’re going to see it more and more. And all personally prefer to manage projects where a customer sends us a million words in the Trados package and we split it up and we send it as offline packages.

 

Sultan Ghaznawi

So, so John, let me interrupt you there for a moment. You just mentioned something really important that some tools work very well for some translators and some don’t do that, you know, so the current tools and technologies that support automation for language service providers, we just mentioned which one for example, supports MTPE, right? Please share your thoughts on how open source is propelling this movement. Do you see that that’s contributing in solving this problem? Because that’s community driven, right?

 

John Moran

Yeah, I mean so that’s a question that somewhere on the Internet there will be a picture of me wearing a T-shirt with OmegaT on the front and somewhere in my bedroom. The the uh… I wore it to a MemoQ conference actually, I think Peter Reynolds took a picture of me… so so when I was working with Welocalize we developed this version of OmegaT that measured translator productivity, we actually didn’t… I wasn’t aware of it at the time, but we were using a very similar method to the method that IBM had used in their TM2 tool. I won’t go into the details of that, but what I will say is that one of the outcomes of the project I was driving in my car with one of the with the girl that I had studied computational linguistics with Jennifer Foster, who was, uh now a professor in DCU, and she told me oh, I have a budget for us to invite presenters and I said, oh, why, she said, would you like to invite a presenter… and I said oh, I’d love to invite this guy called Didier Brill, who was the product manager for OmegaT so we brought Didier over to Dublin and set up a meeting with senior management with Welocalize and I suggested that they integrate or OmegaT into Globalsight, which I know that you have and you were on the Globalsight board and so I was pretty excited about that, I thought, oh great You know, global sites, open source and OmegaT is open source and this will just, you know, ignites the whole open source industry.

 

And I was pretty familiar with the Okapi libraries and Yves Savorel’s work and you know, I’m a huge proponent of open source personally, because the advantage of open source of your programmer is you can, you can adapt the software so it does what you want if you know how to program, so yeah, I’ve been pretty excited. I was very excited about open source for a while. But you know at a certain point Welocalize looked at the economics of global size and decided actually maybe they are not going to support it as much as before. OmegaT has still been doing very well and we still use OmegaT on projects. I think it’s probably the most stable CAT tool and certainly one of the most It doesn’t freeze, you know all the CAT tools will freeze if you put a large file into it, and Omega C is actually pretty performant. And so there’s lots of advantages of open source, but in general, do I see a trend towards open source in the industry? I would say no, and Welocalize paid Didier for his work and but I think that was the largest paid project he had for for OmegaT, he told me.

 

And so it’s hard to get companies except for Okapi and which is kind of a back-end processing framework you know with filters for very good filters for it’s had millions of euros actually invested in into the Microsoft Office file filters. With the exception of Okapi, I’m trying to think, is there any open source software other than I don’t know, maybe like FTP programs or Apache or something, and I don’t, I don’t think that open source is dramatically trending up at the moment, but it’s not, you know, nor is it history.

 

Sultan Ghaznawi

So let’s let’s talk about something different. Still focusing on translation companies trying to somehow leverage automation, right? So when it comes to quality pricing and vendor management of machine translation and post editing, there is some sort of a disconnect between expectation and reality. Interested to hear from you, John, about finding that sweet spot that works for every language service provider or most of them.

 

John Moran

I mean the problem is really pricing, you know, like we are not a translator, I tend to think of us as a translator with a cape. In other words, you know, we can, we can translate, you know, 70, we can translate the equivalent of what one transit of what 60 translators or 70 translators can translate, right? And but we still get you know, emails all the time from our customers that say, what do you charge for post editing? And I have no choice but to say, well, we offer a 20% discount, but that’s not really true, right? I have to see the text I have to see the machine translation output over the years because I’ve dealt with it so much I have a kind of a sixth sense. You know, I can look at it as a translator and say, OK, you know, the quality. We’re pretty good there. Yeah, we can offer that at that particular price. And it’s what I call it you know sort of try and see economics and as opposed to the tools should simply be telling us how much faster we are with machine translation so that the pricing can become a science rather than an art. So that’s my, that’s my pet peeve and that’s really what I was trying to say in the GALA talk that you reached out to me afterwards.

 

Sultan Ghaznawi

About right and when it comes to tools, whether it’s open source or not from inside the industry or from outside industry, remember that Google and others are actually creating tools of their own that that supports translation. Do you find it useful to automate the work that we do in our companies if we were to use something that you know, that’s not created for us or or do you find that that that tools that are customized and created within our industry provides better value?

 

John Moran

Uh, I think tools that are provided you know that are public, you know that are that are from our industry provide better value. I mean CAT tools are the obvious thing. An ERP systems, I think, I mean I know I I certainly know people who are using ERP systems that are more general. But that’s mostly because they implemented them before XTRF and Plunet etc. came along. And we did have an extra XTRF system, as I told you, because I bought one, but we ended up migrating to a system that I designed. We use a Java development framework called VAADIN which is not developed for the translation industry. It’s just a general programmer sort of nerdy framework. Uhm, but I mean other than the sort of trivial caveat of things like FTP programs that every, you know people across all industries. I think generally speaking most of the software that we use is either developed by us for our own business or you know there are things like CAT tools or a lot of the time we’re just logging into an online CAT tool that a customer is provided and sometimes automatically sometimes manually. Uhm, I don’t think that there is a single answer to that question.

 

Sultan Ghaznawi

Now let’s let’s discuss what the users of technologies besides translators they think of the technology itself. I’m talking about people like project managers. If we were to automate many of the repetitive tasks like job creation team, leveraging file preparation, translator, notification and so on, what can they expect that their work will look like at the end once we have something like this implemented?

 

John Moran

Will they have work?

 

Sultan Ghaznawi

OK. Well, that’s a question.

 

John Moran

OK, so I I remember talking to Michael, meet Mikhail Schneider, he’s the [indiscernible], I wrote that down earlier, I just called him [indiscernible]. I remember him joking at TComm that I think I can’t remember how many employees he has in his customer in his company, but probably in the three digits, yeah. So it’s quite a large German agency not as large as something like Lionbridge, but certainly large by German standards, and I remember him joking that in his mind he imagines [indiscernible] just being one person and everything else is automated. It’s one way of thinking about things, and I can see from his perspective running a much bigger business than mine and how it makes sense to think in that sort of uhm, I suppose thought experiments way and I think that most of the steps that you just mentioned there are done manually by us and the reason is because even if you’re doing, let’s say you’re doing project creation and you automatically create the project well, what happens if you’re translating from German to English and some of the sentences or in English, well, you don’t want the translators to get paid for those sentences. You want to copy source to target and then lock the segment. That’s a manual process. I mean, you can imagine it being automated because language identification is pretty much a solved problem in computational linguistics but that facility is in no CAT tool or system that I’m aware of and hint, hint to any of the guys any CAT tool companies out there and there are lots of features of capsules that are not implemented yet. But yeah, so so there are reasons why many of those tasks are manual in my business, in Transpiral, so I don’t I don’t necessarily see it. It doesn’t bother me if if if a project takes 10 minutes. It doesn’t bother me if a 10,000 word project takes takes 10 minutes to create the project let’s put it that way.

 

Sultan Ghaznawi

So, speaking of Transpiral, it’s an SLV, but it’s also now dealing with Irish. Does automation work well for SLVs? I mean, and if you to put things into context, do large organizations with deep pockets profit from deep automation or anyone can implement it and reap rewards?

 

John Moran

It depends, I mean, I think if you’ve got a system like XTRF it depends whether you’re programming or configuring, so if you’re if you can configure a system like Plunet or XTF and you know how to do that, you know you’re going to have to have some sort of budget for that, but because it’s hard to figure out how to do it yourself in my experience. Yes, I think smaller customers can definitely benefit from that. And now if you look at a large agency like Welocalize or even Bayu they’ve got, they’ve got, you know, teams, large teams of programmers, and it’s hard for a small company to compete with that, with that kind of you know, programming capacity. So yes, I mean, of course, automation is open to everybody. I mean, there are some things that we automate, like, for example, we’ll create an extra e-mail address and we’ll give the translator access to that e-mail address.

 

And that a lot of the time that that helps us with what we call spinning plates projects. In other words work that comes in every day and where we don’t necessarily want to have any touch time from the project managers and then at the end of the month you know, we’ll do our invoicing. You know, we put together the work order for them. That might take 20 minutes. We automate where we can, but you know the other option, and I’m sure you’re familiar with it, is that you can you can have project managers in lower cost countries, which is what we do and to do the the manual work that it just isn’t worth automating. And you know, we’ll do that for maybe a few months then once the project is settled down and we reckon it’s worth it’s worth our while, then we would start to invest either my time it’s programmer or one of the team in India. So it really it really just does depend on the, you know on the business case.

 

Sultan Ghaznawi

So speaking of investments, what type of investments do you need to maintain and keep up with developments that affect automation? I mean, we’re not just talking about automation and translation execution, but we are also talking about production and you know, whatever else core functions you have in your business. What are your thoughts on this?

 

John Moran

Yeah, I mean, it’s hard to come up with a number. I was just thinking about the answer to that question, you know, is it 5% of our gross margin or is it, you know or you know some percentage of sales, I, I mean certainly we do have a budget for, for technology. I think we probably spend about €20,000 a year on technology and then maybe €20,000 a year for, you know, going to go into conferences and learning about technology. We have a few consultants on the, on our, you know, on our, on our books that know let’s say Trados better than I do and even though I I know Trados pretty well and I think one of the problems that I think that agencies can have is that they don’t get too caught up in technology. And I’ll give you an example. So going back to the example of a E-teams and I had a very good relationship with the former owner of the company. She was very kind with her time. I remember offering her €100 an hour to help me after she’d helped me too long and I started to feel guilty and you know she, she continued to be to be very helpful and so we had a lot of long conversations about XTRF and you know her approach to the business and to me it felt that she had got a little bit too tide up in the technology side of the business and wasn’t as aware of the fact that the market for Irish was going through a significant dip caused by the financial crisis and over the fact that she they were getting too much piece meal work from the MLVs she was working with one of which was a company I was working with at the time. And basically their project managers when you factored in the cost of the project manager that was eating up all of the margin on the on the work. So she would have actually been better just to focus just on Irish and just you know, let go of the project managers that that that weren’t it wasn’t the project manager. It wasn’t the customers fault, it was just the business and or at least talked and talked to the customers and say, look, we’re losing money on your accounts can you do something about it? Send us bigger projects? Basically, and she did, it’s not to say that. I think she went from offering 50 languages into English down to three. And we do, I mean we we’ve kind of started to make. I don’t want to say the same mistake, but we do provide maybe 10 languages into English at the moment. So you have to be careful not to get caught up in the technology side and forget about the business side. And I’ll just one one comment and this is kind of a little bit of a pitch. One of the one of the the things that frustrated me most about that particular time was I couldn’t help but think back to that five minute conversation that we had in Limerick after Dionn Williams presentation and fantasized that she had said let’s have a cup of coffee and and then she had been completely open with me about the problems in the business at that time because I think if she had said to me, OK, John, I completely trust you wich of course probably wouldn’t have happened. But, you know, even though you’re a competitor, I, you know, help. Exclamation mark. Exclamation mark. And at the time I had, you know, money on a bank account. Well, not that much, not as much as I have now. But you know, I would have been able to help a little bit financially. I would have you know, been able to maybe come in as a partner and I I think I would have been able to say, listen you just need, it’s a bit like I’ve been watching… oh what is the name of that blonde haired cook. He goes in and fixes restaurants that are…

 

Sultan Ghaznawi

Gordon Ramsay

 

John Moran

Gordon Ramsay I’ve been watching all of his episodes, at the moment I just before COVID… I ended up… I had two meetings with an agency in Spain and I was going to come in as an investor with a fairly significant amount of cash and to try to help this business that had gotten into trouble because I was so frustrated by the fact that I had only come along six months after the liquidation of this other company, so you know if there are any agents, any, any translation agency owners out there who would like to contact me… who, you know, need investments or even just an ear… you know, a friendly ear and talk about ’cause not every agency you know is always doing well. Right? Yeah, so I’m at the moment, I think one of the things that I’m hoping to do in the next couple of years is to finally actually work, as you know, even on a pro bono basis initially to try to find, you know, a good match for Transpiral, oh, you know, as a partner or even just, you know, owning two companies from, let’s say, somebody who’s retiring and where I can, you know, use my skill as a, you know, as a as a as a general man and maybe making investments, so maybe some of your listeners fall into that category.

 

Sultan Ghaznawi

Absolutely. I’ll encourage them to get in touch with you. So what does the future of automation in our industry look like to you, John? Where do you see it headed, and do you still see a place for organizations that decide to do things manually?

 

John Moran

I mean, I think there will always be a place for manual work at the start of a of an engagement with a new customer, I don’t, I don’t see, you know, I’ve noticed that academics have a tendency to think of this kind of flying car future, where everything is completely automated and I I think that one of the problems that we’re facing is that people are not encouraging their kids to go and study translation because machine translation is sort of, you know, in German they say in [indiscernible], you know, it’s well known about, you know, amongst the general population and so I suppose the slightly unexpected answer to your question is I think there’ll always be a place for manual work.

 

Sultan Ghaznawi

Yeah, as we reached the end of this interview, John, I would appreciate if you could share your thoughts and advice in this area for our peer executives, other company owners as you mentioned that are on the edge when it comes to automation. What would you tell them to do differently?

 

John Moran

Well, that’s very easy and I see agencies working all the time with us who have no interest and I’m repeating myself ’cause I said this earlier in the podcast, they have no interest in what we think of their software and but often we know exactly what needs to be done to improve the software. And so I think that particularly you know if you are an agency that has their own CAT tool, for example. You know create a… most of the CAT tool companies, have a board, which is people who volunteer their time to give feedback and it’s more than just listening to users or asking people to post feature requests to a, you know, a GRS system or you know an online, product management system, it’s it’s, it’s more about looking at where like what’s the big picture problem? For example, the case of the CEO that I met at Slator conference, and I could have solved their problem had they engaged with me. He actually said yes, and then I got sort of lost in meetings with his employees and so it never led to anything, but the point is I could have really helped them simply by showing them how to push the log output of the desktop based CAT tool that they were using up to there their back end systems so that their support staff could see it, so stuff like that really just engage the users but don’t be myopic about who the users and your customers are you sorry you SLV suppliers are also users, and your translators are users and I think people are always happy to be asked their opinion.

 

Sultan Ghaznawi

Well, what a fascinating conversation, John. I’d really enjoyed speaking with you and I’m glad we were able to have this conversation. I want to thank you for your time and look forward to doing this again in the future.

 

John Moran

Thank you, Sultan, it’s been an honor.

 

Sultan Ghaznawi

Our industry has gracefully adopted automation in the form of machine translation, but also in other areas such as productivity management, sales and much more. Our businesses rely on automation to accomplish their objectives in so many ways, while automation has been around for a very long time artificial intelligence makes it possible to automate a lot more of our work using past human experience or patterns. This opportunity opens up our industry to a world where repeat and boring work is delegated to computers with massive scale and processing power while humans do things that are interesting for them, such as building relationships, performing meaningful activities such as management, and more. We do not need to be scared of automation. It is part of our life. From traffic lights on the road to using Siri and our phones, we have embraced automation spotter for life and we should allow our businesses to leverage this capability as well.

 

That brings us to the end of this episode. I hope you enjoy listening to my conversation with John. If you have any comments or feedback, please share them with me. I promise I read each message.

Don’t forget to subscribe to the Translation Company Talk podcast on Apple Podcasts, iTunes, Google Podcasts, Spotify, or your platform of choice and give us a thumbs up or five star rating until next time.

Thank you for listening. Make sure to subscribe and stay tuned for our next episode.