A buyer’s-side view of localization needs, the decision process, roll out & more
Webinar Transcription
Adam Asnes:
Hello there. This is Adam Asnes with Lingoport, and thank you for joining us for this webinar. I have our special guest, Luz Pineda. She’s a senior localization technical project manager at Avigilon, which is a Motorola company. And she is one of the primary people involved with setting up continuous localization at her company. And so she’s gone through this whole effort. Now, you’ll have to forgive me because the title of this webinar is a marketing term that I’m very familiar with, and I got feedback that not everybody else is familiar with it and, “Hey, is this a sales pitch or what?” No, it’s not that.
A Buyers Journey is like, think of it like a hero’s journey. There’s a hero’s journey when you write an epic, and it usually starts with somebody who wants to try things out, has the vision, goes and undertakes this effort, has setbacks, has trials and tribulations, and comes through and is successful in the end. That’s like every superhero Marvel thing ever. Every big epic ever has that kind of a theme, Disney stories, all that. But what we’re going to talk about is the reality, it’s a little bit more pedestrian, of continuous localization, what it means for us, specifically for software, and all that.
- What that decision was like?
- What were the obstacles?
- What was the trial and how did Luz’s team get through that?
- What’s going on now?
- What’s the vision of the future?
So that’s the summary of what we’re going to talk about. We’re going to do it in a Q&A type of environment. So that’s my brief introduction.
When I talk about continuous localization, just as a point of reference, what I’m talking about is that software, specifically software developers, have their own set of ways that they’re working with that is different than your marketing department in nomenclature and the mechanics of how it all works automatically. It’s unique to software but the idea is software developers are pushing out features all the time. We’re pretty much gone from the long, monolithic, development cycles, where there’d be like one to two releases a year. Now, it’s constantly new features being released. And that’s what we talk about with agile development, this whole lockdown of that, but it happens here.
The idea for continuous localization is that when the software developers make changes that need to be localized, that’s automatically picked up, send out for translation, it automatically comes back and automatically back into the repository. Sounds simple but it’s not because there’s whole bunch of variables in any organization. There’s always exceptions. Maybe this team does it differently this way or that way. And that’s the problem. And the solution is not necessarily obvious, and it’s not a one stamp. So we’re going to talk about how do you make that work, practically, in real life.
All right. So a little bit about Luz. Luz has been involved on the technical side of all the interactions with Lingoport, and in particularly, with Oliver Libouban who many of you know from other webinars. He runs product development here. He likes to get really involved with our customers. Oliver is actually standing by. He’s not on the marquee of this webinar, but he’s there in case you want to throw a zinger. Oliver is there if you want to ask something technical that’s above my knowledge base.
So anyway, Luz, in your own words, can you talk a little bit about your role at the Avigilon and Motorola?
Luz Pineda:
Absolutely. Yes. Thanks, Adam. Well, I just started my career in Canada working as a localization tester for video games. I know. It sounds fun, a lot of work, believe me. Then I spend over eight years working in different roles in software testing, and I have been working for Avigilon for almost six years. This Saturday is going to be my sixth anniversary. And I work for the localization and technical documentation department led by Paula Hunter. Avigilon has a centralized localization program that supports the entire organization. Of course, we focus on localization for software, technical documentation, website, marketing materials, and customer experience.
I was hired to focus on supporting the engineering team, so for localization. Of course, soon after I started working for Avigilon, we also started a Co-op intern program in order to add automation to our processes. The initiative was actually a success, to the point that we were able to convince one of our former interns to join the team permanently. Her name is Jane, and you are going to hear me saying her name over and over because she’s been all the way helping us onboarding more products softwares into Lingoport. I recently got promoted to Senior Technical Localization Project Manager, long title, and my main responsibility is to drive the software localization operations.
Adam Asnes:
All right. And in praise of interns, we have them too. They do all kinds of special projects that are really cool. Anyway-
Luz Pineda:
They are great.
Adam Asnes:
Okay webinar viewers so this is where I’m just going to ask Luz questions, and she’s going to answer.
- How is your continuous localization rollout going so far?
- Paint a picture of where you’re at?
Luz Pineda:
Well, considering all the moving pieces involved in integrating both tools, Lingoport and XTM, I would say we’re making great progress. It has not been easy or problem free, believe me, but thanks to the excellent support that we have received from you guys and all the teams involved in this effort, I would say that we have on boarded over 90% of all the repositories that we need to monitor for UI changes.
Adam Asnes:
All right. Very good. All right. This is always interesting to me. Where does localization sit in your organization? Do you report to development or marketing or something else?
Luz Pineda:
Actually, the localization team along with the technical documentation team reports into the product management organization. So and as I mentioned before, we support the entire organization, focusing also for development. We have 18 repositories so far that we need to constantly scan. Our local set goes from two languages all the way to 23, so there is a variety of language support. We also support the technical documentation team, which means we localize user recommendation, spreadsheets, quick style guides, inbox items. On the side of marketing, we localize the website, localize marketing collateral materials, emails, forms, campaigns, so many, and the customer experience as well, end user and partner communities.
Adam Asnes:
So would you say you’re spread thin? You’re not overstaffed?
Luz Pineda:
Trust me, we’re not.
Adam Asnes:
That’s another thing in common with every localization too. Okay, good. All right, so before you had any continuous systems in place, maybe describe what was your situation and requirements and what was going on in your organization.
Luz Pineda:
When I joined the organization, it was busy, but it wasn’t too bad. I was basically just scanning five repositories. I was well aware of all the milestones and dates. It was busy, but it was manageable. However, the company has been growing significantly in the past couple of years, both organically and through acquisitions. So now the number of development teams has increased, of course, as a result, both locally and in other countries. So we now have geographically dispersed software development teams. This includes countries such as the United States, Poland, Scotland, just to name a few.
Now, we have more features that need to be delivered faster, and we have more markets to reach, also as a result of this growth. Now this last one, more markets to reach, as you can imagine, it means more locales need to be supported by our software. Furthermore, our cloud-based solutions are adopting microservices architecture approach. And if you are familiar with microservices, everything is independent. They can release anytime they are ready. They can deploy anytime they are ready, and they also need to be localized if they are going to be used by other solutions, which handle a particular language set.
So over a little bit of a year ago, I would say, it was clear that unless we found a way to automate some of the localization workflows, we were just not going to be able to keep up with our development teams. It’s not a matter of just throwing bodies. You can add more project managers, but that is not really scalable. You can make clones of Jane or myself, but again, how many clones do you need? It’s not scalable. So this is a situation where the proverbial do more with less applies completely. We really needed to do more with less, given the fact that the number of repositories more than tripled in the last year.
So there is also a need to increase the release currents. This aligns with what you mentioned earlier. We need to release faster if we are going to keep on top of the market. Our previous processes unfortunately, did not offer that flexibility. We needed to deliver translations in a timely manner, but we also need to keep localization affordable, and that was not the case. So that’s basically when our quest to find the right tools started.
Adam Asnes:
All right. For those who maybe are new to the software world, when we talk about microservices, it’s a specific type of software architecture that really isolates different features that can be plugged and played in differently with different products. And it just, for localization, think of it this way. It adds complexity because there’s more of it that has to be tracked, and it’s never moving in parallel. So one team is doing this, the other team is doing that. They have a very different cadence, and so things like visibility to what’s happening there, automation become more and more important.
So in some ways, continuous localization is friends with increased complexity. The two go hand in hand often. And microservice is this whole movement that we’re seeing across the board in development teams. So I have to say, it’s extremely common that people are moving to that, and it’s causing new issues for many localization groups. All right. So take yourself back. What did success look like when it started?
Luz Pineda:
We wanted to find a tool that you press a button, and everything was done automatically. No :). We were actually trying to find a tool that definitely scanned repositories on a regular basis. That was huge. As you said, things get done not necessarily in parallel so you need to be on top of things. We also needed to be notified when changes in the resource files happen and be able to automatically send those changes for translation.
That was something that it was important for us. And to be quite honest, at the time, we were not sure that the same tool was able to automatically download the translations once they were done. We didn’t really know that part that much. All we wanted to do is, okay. So there are new things in the repositories. We need to know when that happened, and we need to send them right away for translation. Once the translations were done, we can … Okay. So we’ll find a way to deal with that. But we actually found Lingoport, and Lingoport downloads the translation, which is great, and commits them in the repository directly, which is also huge, because all you have to do in the morning, sit down with your coffee, and just go over the repositories and make sure as to what you need to merge into the target branch.
We were basically looking for a process that was as hands free as possible. That was the main. How that process looked like at the time, we didn’t really necessarily know. And we also were very keen on keeping our cost affordable because we didn’t have neither the human power or unlimited money to spend to keep localization ongoing. So that was basically what we wanted, something that finds strings for us, let us know that there are new strings, send them for translation. And on top of that, we found something that once the translations are done, puts them back in our repositories, which is huge.
Adam Asnes:
All right, great. So security, I know that came up big. When you have a continuous localization system, by nature, you’re connecting with a software repository, where the source code is. So maybe you can talk about that hoop, in your words, rather than mine.
Luz Pineda:
Yeah, absolutely. So source code is one of the most important assets for Motorola, no doubt. They connector we chose have to be installed on prem, and so that means any cloud-based solutions were completely out of the question. And we have to prove that the only resource files that were sent out to our translation partners were only resource or language files, nothing else, nothing even resembling source code. And at the same time, we have to prove that whatever we got back from our vendor were only files with the strings, nothing even remotely that looks like source code.
So that was something very important for us, and at the same time, that meant we have to evaluate both tools, Lingoport and XTM. They have to be thoroughly tested by our IT department. They raised concerns, they raised questions, and we were able to address those concerns, Lingoport was able to address those concerns, in response to those questions in a timely manner because this process, when you say it, it sounds fast, but it’s really not. We went through a series of weekly meetings with the security team, just talking about what the progress was in this testing and what the concerns were. The measures to address those concerns have been taken of note, and to make sure that they were happy with what we were going to install. And that’s how we basically were able to overcome the security step in this process.
Now, on top of that, of course, our big issues have an intuitive user interface. Because we were going to be ultimately the responsible for maintaining the tool and maintain the daily activities. No external consultants or integrators were going to be allowed to access neither the server or repositories. At the time, there was a big chance that only one person, myself, was going to be able to manage everything, maintaining the tool and keep up with daily activities. So I didn’t really have the time, too much time to spend in learning how to operate the tool while keeping up with my normal workload. So it has to be something relatively easy to use. Luckily, Jane join us, and now we are both basically up to speed on how to manage this tool, to manage Lingoport.
Adam Asnes:
All right. I’m just going to mention that security is just … Dealing with security concerns happens at almost every one of our customers. Yes, all of our products do touch source code. In the case of Globalyzer, it’s literally scanning what the developers are doing daily. Localyzer also is always scanning repos. So security has always been super important. And while we do offer hosted services for those that want it, I would say it’s an 80/20 rule. Almost every one of our large customers goes on-prem, and we understand that. It’s a sensitive thing.
All right. So you didn’t just jump in with both feet. You ran a pilot. So that’s part of your journey. Tell us about that effort.
Luz Pineda:
Yes, of course. Yes. So as I said, we started looking into tools more than a year ago. So it has been a slow process, but we just have to make absolutely sure that what we chose, it was what we needed. So for the TMS evaluation, we select the software products based purely on file format. As you can imagine, having 18 repositories, it means that we handle every file format under the sun. For Lingoport, we actually selected software products based on the priority. So we used our flagship products for the trial with Lingoport. And, as it happens, those flagship products are actually handling very challenging file formats, and Oliver can attest to that.
Of course, some customization was required, and we have to make sure that Lingoport could address that kind of customization to the level that we needed. So as I said, we evaluated two TMSs. We evaluated Lingoport with another connector. We weighed what we needed and what we wanted, so to speak. We assigned weights to those nice to have or need to have. And that’s how we actually chose Lingoport as a connector, among other things, of course.
Adam Asnes:
All right. You bring up the issue of file formats. There are some file for files like properties files that are … Java properties are pretty well defined. JSON, for instance, is a little bit of, I’ll use an American youth colloquialism, it’s the Wild West. All JSON file resource files for translation will have this format. So Luz is talking about the need … It’s a lot easier if you can be flexible in terms of what comes in and out, versus trying to get the developers to change, which is a battle. We’re going to talk about stakeholders in just a couple of slides.
Oliver Libouban:
It’s Oliver here. Luz, can you talk about some of the file formats that we did together?
Luz Pineda:
Yeah, absolutely.
So we have at least one software product that uses a customized solution to implement the localization. So it’s a txt file. So yes, of course, you need to transform that into something that is localization friendly as something that has to be consumed by XTM. We also have YAML files, and we’re still having a lot of fun with those type of files. Remember that at some point, I mentioned that we should be able to send only deltas. So grabbing those deltas from a YAML file has been a challenge. We’re still working on that, and I will talk a little bit more about on the development side, the efforts that we’re trying to actually change that file format.
We have XML files. We have JSON files. Lucky us, we were able to get inserted into the microservices development teams very early. So we are using JSON files for all the microservices, and it has been great because basically, we’re using the same format for a good chunk of software products, and we have property files, we have RESX files. Am I missing anything else, Oliver? I think those are the main file formats that we’re using. But yeah, definitely, we need some transforms in order for XTM to consume those files.
Adam Asnes:
Okay, good. And I told you Oliver was listening in. He was like the voice from beyond there. Let’s talk a little bit about budgeting. Every organization can’t just take out their credit card for enterprise software. There’s quite an impact. Even outside of money, there’s impact on the organization. So maybe talk about that process and what you were measuring for and that sort of thing.
Luz Pineda:
Yeah, absolutely. So well, there is obviously a cost attached to this tool. However, the time savings that we’re getting are significant. And sometimes time, you cannot even measure. Well, you can measure it in terms of money, but time is something that either you have or you don’t have. So at this point, we’re at the point that we don’t have to spend hours preparing translation packages, which it used to happen before. And there is no manual process of the resource files in order to identify the changes in such files. We don’t actually need to QA the translations deliveries on their way back because this is done by Lingoport.
That is also a significant amount of hours that we are saving just there. We have the confidence that we’re not missing any new strings, and this is very important to point out. Because if you know that a team is going to add 20 strings today, you go into the repository and there are only 18, you still need to send that translation because they are waiting the translations back, I don’t know, tomorrow or the day after. You cannot wait for two strings to come and then send the whole package. But by now, we can actually do that. We can send 18 strings, and we can send two strings tomorrow. And that is okay because we don’t need to keep on top of those changes ourselves.
Additionally, we have completely eliminated the cost and time spent in pre-processing and post processing each translation request, huge as well. That means basically, we don’t have to wait for all the 23 translators to complete their assignments for us to start merging translations into the target branch. We can just start merging those translations as we get them. There is no post processing anymore. We can add more repositories, more languages without the additional complexity that sometimes this brings. And we didn’t have to spend any time or resources that we don’t have developing or maintaining our own connector. That is also huge.
It sounds appealing when you know that there is an API that you can use to create your own connector. But it’s not just coding for the new connector. You have to test it. You have to maintain it. If something goes wrong, you have to troubleshoot on the spot. Otherwise, the whole automation of the translation workflow is not there. It defeats the purpose.
Adam Asnes:
All right. Thank you. I always like to talk about stakeholders in an organization, because it’s never just the localization team. You have your internal customers. Maybe talk a little bit about your concerns and experiences working with the software development teams, as well as localization vendors. And I’ll just preface that it’s really common for that relationship between localization and software development to be a little thorny, actually, or difficult for some people on the localization side.
Luz Pineda:
Yeah, yeah. You know what? It sounds made up, but no. We’re actually really lucky to work with an amazing group of software engineers. They actually get localization and they know that it’s important to keep it in mind when they are doing the job. I know it’s unheard of, but we’re just lucky. And maybe because we’re fully embedded into those teams, you get to know them at a personal level as well, which is important.
So we have developed this mutual relationship of trust and collaboration with them. It was very important that the transition from the previous tool to the new tools was as seamless as possible. You need to maintain that relationship. And basically this meant that we needed to be able to continue delivering translations while the transition took place, and keeping delays to a minimum, really to a very, yes, the minimum. They shouldn’t even know that this was happening. And the way that we approach that is we talk to each of the repository owners, and project managers of each team, because you need to talk to them in order for them to grant permission to Lingoport to access the repository. So they knew this was going on, but the rest of the developers they didn’t even know.
And that’s exactly what we needed. Because otherwise, it’s like, so what’s going on? Why we’re not getting the translations when we need them? On the vendor side, we created this plan. So they knew in advance the order in which we were going to award every software project. This helped them to make sure that the templates were created in advance, the parcels were configured, workflows were in place and set up. And most important, that the translations memories were uploaded in XTM before we send the first translation round, because we still needed all the leverage from those translation memories right off the bat.
Throughout this process, of course, a big shout out to the Lingoport support. Oliver has taken us by the hand, step by step during the whole onboarding process, creating transfers for us, helping us troubleshoot whenever issues arise. Because as you can imagine, something is going to happen. And I think it’s better to know that rather than oh, okay, so everything is going to be okay. Nothing bad is going to happen. No, trust me, something is going to happen. So it’s really helpful for us to have somebody we can rely on and we can ask questions, whenever we need. I know it’s also unheard of, dealing with customer support. We get answers to our emails, most of the times, in the same day. So that is huge for us. So thank you so much for that.
Adam Asnes:
All right. Hey Oliver, I know you’re listening to that, and I hope you’re turning red. It’s good.
Luz Pineda:
Yes.
Adam Asnes:
All right.
Luz Pineda:
I did that on purpose, yeah.
Adam Asnes:
All right. So let’s think about scaling up. What concerns and what visions do you have going forward as things invariably get more and more complex in terms of the organizations that you support?
Luz Pineda:
Yes. That is actually a good question because we know things are not going to remain the same. The only constant is change. It seems in our case, or the case for the company, the only change is to keep growing. That’s something that we need to keep in mind. So although the onboarding of the server products has been relatively fast, it has not been completely without bumps on the road as I mentioned before. At the risk of sounding overconfident, I think by now we have a good grasp of what could go wrong and most important, how to fix it or either work around it.
So I guess my main concern is about what we don’t know. Earlier, we talk about the growth of the company also through acquisitions. So there are definitely development teams that are joining us that do not know the way that we do things. We don’t know the level of localization or internationalization knowledge that they have or they keep in mind. It’s still an afterthought in most development environments. It’s, “Oh, localization? Yeah, we can do that later.” It’s not the case.
So I guess my main concern is, as we get more products onboarded, and we have to help more development teams, is what exactly are they using? Are they using repositories that we haven’t tested before? We haven’t tried before? I think those are my main concerns, the things that we don’t know, because the things that we have experienced, we know how to fix them. We know where to start the investigation. But yeah, there might be things that we don’t know and that would be of some concern. Of course, if we keep the same level of support, I can sleep better at night.
Adam Asnes:
All right. Good, good, good. All right. So we have already a lot of questions. This is where we’re going to move into Q&A. We have a good amount of time for that, so that’s great. We were good with our allotted time. Thank you Luz. Your answers have been really crisp, really good. All right. Is the budget for localization centralized or do you require different departments to fund localization projects?
Luz Pineda:
Well, as far as I know, we have, within the localization budget, we have different buckets. So one bucket is for the server localization, another bucket is for marketing, and another bucket is for technical documentation, for example. So whenever we get a quote, it goes into a particular bucket. And it’s Paula really, the one that so far, she’s managing all the budgets. We provide input, of course, at some point during the year as to what we expect for the next year so she can start planning. But really, it’s just that kind of division, software, documentation, marketing, customer experience.
Adam Asnes:
Okay. Very good. This next question, I have my own answers to this, as what I see as the future of localization. But anyway, this question is, do you use MT for translation of UIs pulled from the repositories?
Luz Pineda:
No. We don’t use MT. We do use MT for technical documentation, but not for software, which is a leveraged with whatever we have in the translation memories, and everything that is not 100% or a nice match has to be translated from scratch.
Adam Asnes:
Right. All right, all right. One of the things where this question may have come out of is one of the things that I have personally been throwing out there, is the idea, and this is not used by Luz’s team, the idea that localization should be always available to the development team. So it’s always updated via MT and then you clean it up, of course, using one of our products, Localyzer QA that lets you go through, let’s say, the latest feature in context and actually see it in the screen. And right there, correct the translation.
So if it’s so fast to do it in context, you can actually get your QA adaptations of the language done right there. And now, you’ve eliminated one cycle of you’re sending it out for translation, and then your QA, doing the linguistic QA. You get them both done it at once, and then the developer always sees, let’s say, their feature, they can toggle between locales on their settings and always see what it looks like in other languages, so you have that full visibility round trip.
So that that is a possibility for teams that I see in the software localization’s future and that we’ve tooled around that. Me saying it and the marketing accepting it are two different things, but it is possible. And I even wrote about it a little bit on our blog, thinking of localization, like a utility, like electricity. You don’t question that you plug a lamp into the wall, and it turns on right away. You wouldn’t expect it to take a day or two or three to get the translations, to get the electricity back. That would be bad.
So I could see with the technology, MT has gotten a lot better. So if as long as it’s easy to clean up the MT in the software, why not use it? And that’s not a question for, I don’t want to put Luz on the spot for that, or make you feel like, “No way. I don’t want to do that. What if I mess it up?” And I get that, there’s adjustment that’s needed for that. So All right.
Next question. Continuous localization can come … By the way, I haven’t previewed these. Continuous localization can come at a cost with minimum charges and project management often imposed by LSPs. This goes against the goals of continuous localization, where small batches are frequently sent for translation. How does your current LSP perform in this regard?
Luz Pineda:
So that is one of the things we’re working on right now. We are trying different approaches to tackle these. Right now, the main achievement we have done is that we are putting together all the translations sent in a period of time, regardless of the product or the template they are using for translation into one quote. So they basically allows to, instead of sending one quote for ACC, one quote for ACM, blah, blah, blah. We put everything, using the word count for every single project together in just one single quote.
It is true that is an issue because sometimes you might not even send any strings. It could be that for one day or two, all the changes are happening on the server side. There is no user facing strings, so they don’t send anything. And you cannot ask them to have translators just waiting there for you to send translation. That is a reality that we have.
So what we are trying to do right now is maybe on their side, and they have to manage that, is something that we’re not going to do anything about it. Maybe they can put together several clients and say, okay, we can guarantee these contracts or these many strings. So when they receive the translations, at least is not that they are going away for nothing. They know that there are strings, regardless of whether the strings are coming from client A or B. They are going to have a guaranteed influx of work so they can respond better to our needs.
This is still in progress. I have every confidence that we’re going to get there. We have also a very good relationship with our vendor, and they have been super supportive in the whole integration of tools, and they are working really hard on that right now. But yeah, that is a valid concern.
Adam Asnes:
Just because I see a lot of different localization companies that are part of the whole value chain for customers, I’ll mention that there are some vendors that adapt much better to continuous localization than others. That said, I don’t want to say anything on the air here while it’s being recorded, but there are some organizations that really understand that this is what they have to support in order to create the success of their customers. And of course, there’s going to be trade offs.
I will say that the other thing is if the data that comes to the localization vendor is consistently good, the files are consistently good, the minute the localization vendor has to spend QA time on 18 words, that’s bad. And that’s understandable as a problem. This is definitely a case of make sure whatever you do, if you’re taking on continuous localization, that your vendor gets clean files, that they can work out without thinking they’re going to have to sign up an engineer to oversee these files.
All right. Very good. Hopefully that answered that question. I think it’s a really important one. All right. So you use Lingoport tool along with some other TMS, right?
Luz Pineda:
Yes. Yes.
Adam Asnes:
Lingoport, by the way, is a suite of software there. If you go to lingoport.com and hit the Products section, you’ll see that there’s internationalization, there’s continuous localization, end QA solutions. And even within each one of those, there’s components like dashboards, and all this stuff that works together. There’s a bot that works in Slack so that you can onboard projects right from slack. There’s some really cool stuff on there, and it also works with Microsoft Teams. The pride in me bristles at the word tool because I think of it, it’s like this is a product set.
Adam Asnes:
Can you support separate import export workflows for multiple branches of code?
Luz Pineda:
We haven’t tried. I know that it is possible. We haven’t try it. Right now, we just follow the main target branch, and we have this … And this is why it’s important to have a good relationship with the developers. So even if they are working in different branches, at the end of the day, everything related with localization gets merged into the massive main develop branch. So that’s where we pick it up. And if necessary, we carry big translations into other branches. But that is something that we do on our side. We haven’t tried the multiple branches scenario.
Adam Asnes:
Just in terms of what I’ve seen, I can think of one customer with a single product that’s supporting 240 odd repos simultaneously, and that’s just one product in the organization. So sometimes, it can get very complex. So it depends on your organization. All right, about mixing strings, are you using any pseudo localization methods?
Luz Pineda:
No, not for software.
Adam Asnes:
Okay.
Luz Pineda:
Yeah.
Adam Asnes:
I will say that that is automated in Localyzer.
Luz Pineda:
Yes, it can be. I think we tried for the first product that we on boarded. But it really didn’t really take that much time to onboard every project that we decided like, “No, let’s just jump right into each of the translation rounds.” And we didn’t have to do anything like that for the rest of the projects, yeah. But it is possible to do it in Lingoport.
Oliver Libouban:
Just one little thing here is that we automatically detect missing translations, and that is actually what’s being sent to the TMS. So pseudo localization is nice to test the head, but in reality, we handle all the missing translations.
Luz Pineda:
Yeah. I think that’s when I mentioned that we are confident that we’re not missing anything. That’s really something nice to have.
Adam Asnes:
All right. This is a question that I actually didn’t address at the beginning. It says I may have missed the beginning in the beginning, but in your words, what is the difference between internationalization and localization? Luz, do you want to take a first shot at that?
Luz Pineda:
Yes, of course. So basically, internationalization is about enabling your software to accommodate different locales. Once your software has been internationalized, then you start localizing, which is taking the cultural characteristics such as language, culture, you name it, from a particular market, and then implement it in your software. You may have a more fancier way to say it, but it is basically internationalized is something that can accept other locales. Locales is actually work on that specific locale, and implement it into your software.
Adam Asnes:
Yeah. I couldn’t have said it better. The only thing I’d add is that there’s a tendency to, in our industry, to think that internationalization is just about externalizing strings or the words that are in the application, putting them in a separate file. That’s certainly part of it, but you have a lot of other issues that include, literally, the logic of how the software works. Luz mentioned locale. That’s language, but it’s also there are issues like the developer may have broken up a message into programmatic elements or concatenations that build the string over based on variables like plurals and the like. And those might not work in your target language. The word order may be different. There may be several ways to express a plural. There may be masculinity, femininity of your statement. So you need to account for that.
Also, number formats, date formats, currencies, address formats, phone numbers. The list is actually pretty long, and it’s also different. That list is different in different markets.
Luz Pineda:
Yeah, definitely. And there are things like the ones that you mentioned that there are somehow obvious, but there are some other ones that are not as obvious. For example, being able to support utf8 in your databases. So if you have people, not English speakers creating a data, input data into database, that database should be able to support those characters. Otherwise, you cannot use user defined data in your application.
Adam Asnes:
Right, right, okay. Next question here. I am understanding that Lingoport is used for extracting source content from code, et cetera. What is the translation platform connected with Lingoport?
Luz Pineda:
Yes. Yeah. It’s XTM.
Adam Asnes:
Right. We actually connect to quite a number of BMS systems. You mentioned, it’s a follow-up question here, you mentioned the use of XTM several times. Is that a required TM for Localyzer? I’ll answer it, no.
Luz Pineda:
I just got to say no. No, it’s just the tool that … Remember, when I mentioned that we were evaluating, at least two TMSs at the same time, just to have something to compare to? That’s the one that we chose at the end of that evaluation.
Adam Asnes:
Right. But I also say we even have customers that don’t have a TMS. They just have us connect their localization vendors portal. So there’s a lot of flexibility there. Okay. Is Lingoport connection to CMSs possible as well? That’s one of those it depends things. We build around software. That’s our jam, software development. There are cases where a lot of content is being used in Git Repositories. There’s always the chance for something customized, but we don’t really focus on pure content management systems, per se, at least at the time of this recording.
We have so many questions. I’ll keep going, by the way, but we’re we are running out of time. So if you need to leave at the end of the hour, let me just say thank you for attending. You’ll automatically get a link for the recording. It usually takes a few days to get that processed and up on a web page and the like but you don’t have to do anything. You’ll get it. All right. Does the TM sit at Avigilon or is it controlled on the LSP side?
Luz Pineda:
Sorry, the translation memory?
Adam Asnes:
Yeah.
Luz Pineda:
Yes. No. That is on the vendor side. However, as part of the delivery, we get exports from XTM. It’s actually a very good question. I’m one of the people that believes that you should own your translations. It’s an asset. So you should be able to get it in a format that you may use it for something else, and we do. We actually use the UI translations in the technical documentation. So if the technical documentation is saying, “Okay, click on download video.” The download video translation should be the same as the one that you see in the software.
Adam Asnes:
Very good. Okay. For eCommerce … I’m thinking that’s an incomplete question. I’ll just say, yes, we do support eCommerce applications. In fact, I’m not legally able to name them, but the two largest eCommerce providers in the world use our software. Everybody here has probably bought from at least one of them. That’s all I’m allowed to say. All right. MT is an interesting topic, garbage in garbage out. The person named their company, but I’m not going to name it. That’s just … Measured the amount of fully formed strings in the UI and identified that 30% of the UI strings could be effectively translated. How do you see MT evolving in your organization in the future?
Luz Pineda:
Well, I definitely agree. And we make a point to review the translations, even before, sorry, the strings, even before they go into the repository. If the string is ambiguous, is wrong, it has grammar issues, that problem is going to translate in as many languages you’re translating that string. So you definitely have to clean that up. I’m a little bit concerned about using … I wouldn’t use machine translation for our software, given the fact that the terminology that we handle is very industry specific. So we need to use the correct translations if we are going to let other users to understand what we’re trying to say. I guess it depends on your risk tolerance of getting a translation that is completely wrong. In our case, that’s very low. We need to get things right from the beginning. Remember that we’re working for security and video surveillance, access control. We need to make sure that the operators of the software understand exactly what’s going on.
Adam Asnes:
It’s interesting, you say risk tolerance. For instance, if you’re in eCommerce and something would have never been translated anyway. Usually, the risk tolerance is fairly low, right?
Luz Pineda:
Yes.
Adam Asnes:
I’ll also mention that there’s so much going on the machine translation front in terms of the switchover for statistical to other forms of translation engines that have increased the blue score. That’s the score of accuracy of machine translation. Plus some engines are much more trainable than others. Where we’ve been doing a lot of testing ourselves to, like I talked about, the plugging in the repos all the time. Well, some of that depends on the accuracy of the translation.
Luz Pineda:
Yes, exactly.
Adam Asnes:
The translation. So that’s probably another webinar. I actually need to have another-
Luz Pineda:
That is another webinar. Absolutely. Yes.
Adam Asnes:
I had somebody from Systran on about a year ago, talking about this, but it would be nice to get somebody from the Amazon team and the Google team and some others, somebody from STL and the like. So that’s a homework assignment for me later this year. I’ll have that webinar.
All right. That is all of our questions. We’re finishing two minutes after the hour. Thank you everybody. Most of you stayed on. I really appreciate it. Oliver, we didn’t need you for any zingers, but thank you for joining in the background. Luz, you did a great job. Really appreciate you sharing your insight.
Luz Pineda:
Thank you so much. Thank you for having me. It was fun.
Adam Asnes:
All right. That was really fun.
