SB01 Transcript

Ep01 – Challenges of Continuous Localization


Welcome and thank you for listening to ROAR: Speedbumps. My name is Ghio. I’m currently a graduate student at the Middlebury Institute of International Studies at Monterey, studying translation and localization management. For those unfamiliar with the title, we are a student led podcast designed to address speed bumps or challenges that exist in the localization industry. We speak with students, educators and industry professionals to learn about the speed bumps and discuss possible solutions. Today, it’s my honor to bring one of these professionals, my dear friend, Melvin Su. He and I go way back, and in fact he single handedly changed the course of my life some four years ago when he convinced me to move to Taiwan. He currently works on the client side as a program manager at Autodesk in San Francisco. Hi Melvin.

Melvin: Hi, I’m Melvin.

G: Thank you so much for being here with us today. I really appreciate the time, I know you’re working really, really hard. So, do you mind just, you know, briefly sharing a little bit of information about yourself, your background and how you got into localization.

M: That’s no problem. My name is Melvin Su, and I have worked for, for my undergraduate studies I studied linguistics and for my master’s studies later went out,  went on and studied linguistics as well. However, I wanted to do something that’s less theoretical in terms of dealing with languages and do something that’s more functional and more applicable to everybody’s lives, I would say, so I after graduating I looked into different options. I started working for Apple as a sort of like a language specialist position. And then later on went to Netflix for a coordination job in the localization field. And then after that I came to, I went to Uber. As a Program Manager. And then, recently I recently just switched to Autodesk also as a localization Program Manager, working so basically my, I would say my whole career has been dedicated to the localization field. You know worked as a coordinator dealt with, linguistic quality, now I’m working mostly platform and program level stuff for localization.

G: Cool. I mean, actually. I didn’t really mention this before, I did meet Melvin was actually at Apple. We were both hired on at the same time. And I find it kind of hilarious that after even those four years since since that time you ended up getting into localization. And I ended up studying, studying it, and it’s I don’t think it’s ever something we ever really discussed together. I know we both had a passion for languages and I’d always bug you with Chinese questions all the time. But to see that we’re actually like in the same industry, it’s pretty pretty crazy. And that, so. So is there anything else you want to mention about your background or can I start bombarding you with questions on how localization.

M: Yeah, I think it’s pretty interesting I like everyone in the localization field, kind of, you know, came in because mostly we’re interested in different cultural backgrounds or languages, and so it’s. Yeah, a lot of our focus has been placed on the language itself, the languages part itself but actually what I been realizing in the program manager role and localization is that like negotiation, negotiation skills and also professional skills is more important than the like language aspect.

G: Right, right, but you still use your, I mean, from the discussions we’ve had before you, you, depending on the project or with the needs of the company, you still incorporate your language, right?

M: Of course I mean, knowing more, how languages function would really help and, you know, you’ll be more sensitive, in terms of how you manage the programs, but I think what really gives the success in the localization field really is your program management, um, management skills, instead of your language skills.

G: Got it. So speaking of that, the reason you know invited you to do this podcast and discussion was because we’re really curious about, you know, speed bumps or challenges that professionals or students are facing in the industry. So, I wanted to ask you if, you know, are there any challenges that you’re currently facing because I know you’ve been just working on the on the on the client side. Have you witnessed any of these so called speed bumps or challenges during your time in the industry?

M: I think a reoccurring thing that I have witnessed. Is that a lot of these companies, be it whether when I was at Netflix or Uber or right after Autodesk, we’re really focused on mainly three things, right. Three being the cost, how can we keep the cost low, SLA, how can we keep up fast turnaround time fast and then lastly is the quality, how can we do everything else and still not suffer on our quality. So, most companies are moving towards what I think people be somewhere. It’s called CL, which is continuous localization. Yeah, is it’s a mapping, or connecting setting up the pipeline for translation as engineers code new, new strings and new content, and then it just automatically goes into translation to, to the vendor side, you know, once we have the pipeline set up. Everything is continuous, we just need to kind of set the cadence, you know pull it once we and then dispatch that job and then have the translation vendor translate on a weekly basis or according to the SLA. But what I see is that it’s, it’s of course it’s not easy to set up that pipeline but once that’s set up, you always think that everything is is kind of just what you would just be going and then you don’t really need to manage it, but that’s not usually the case.

G: And you probably won’t have a job if that were the case.

M: Haha yea. I think, of course, other than the you know the hiccups in the tools and those issues. What the reoccurring thing I kept seeing that would, I think would be a challenge or speed bump. What that the content just keeps coming in right so that keeps getting these, these new contents to translate, but a lot of times like linguists have very little visibility on what they’re translating so all they see is just it’s just one string and then they don’t know where it’s going to appear and the product, or they don’t know what new initiatives the that the client’s team is trying to do or you know what our company is trying to do. The translators has no visibility, so it is that is getting that context to the translator or to the vendor that has been, like, a challenge for most companies. Not because, not because we don’t want to provide it though because the continuous localization setup is that it gives very little visibility. And often times, the localization team inside a company like on the client side. A lot of times, you can get your part. You know stakeholders to set up these pipelines, but it’s very hard to convince them to include you into their, their design stages or in their meetings where they’re discussing what they’re going to you know how they’re going to how they’re going to design this new product. So a lot of times the localization program manager has no visibility on what’s on what’s going on to the pipeline.

G: So let me just make sure I understand this so essentially you’re getting this content that needs to be localized, right? You’re not really familiar with what it actually entails in the ins and outs of it and then you’re just expected to give it to your vendor who then doesn’t know anything and because you’re so tight with the timelines that there there’s just kind of like this gap where they don’t know what’s happening and but you, the, the other product teams are asking you know we need this localized now, so is that is that what, am I understanding that correctly?

M: Yes, or especially when, when it’s already set up for continuous translation. So, things just go through that the vendor gets the content, and we don’t even know that it’s going through, because the pipelines are already set up, you know, and then they might be something new, then they don’t know what they’re translating, so that’s going to.

G: So what has been some of the best I mean obviously the benefit to that is just the speed, I’m assuming right just, you know, the automation aspect of it. So is the biggest, you know drawback to that the quality? or in your eyes, what is the biggest challenge with that constant localization?

M: I think quality. Definitely suffers. But definitely, this is something we can mitigate, and I think this is where a real challenge. Is this where, like most. These are the issues that program managers should solve, as soon as they are entrusted with a program, I think there are a few things that you can do to mitigate this, quickly reach out to your product managers, even the ones that you’ve already set up the pipeline, localization support does not end there. Even though they can submit content and, you know, everything’s kind of running on its own, but it’s not as without that crucial piece of context or knowing what your product managers are working on your vendor cannot help you perform well. So, I think it’s really important for for us as program managers to set up, you know weekly meetings or bi weekly meetings, just to understand what products or what initiatives these product managers are working on. And just, they think that they only need to reach out to us at the when they submit the content but that’s not how it is you kind of really, really need to know what’s coming. And what is the launch roadmap. When is it going to be live? So, that you can properly prepare your vendor for success.

G: Okay, so in the cases where quality suffers. I know you mentioned to me like it depends on the visibility right, so things that are more visibility obviously the quality is more important. You know where the end users have to deal with it or whatnot, like the website, etc. So, for the from the aspect of the teams that you, you work from from their point of view, what do they understand quality, to be. Do they understand quality in the sense of you know localization localization professionals, or is it just a kind of this very general idea of like language quality?

M: I think, what is quality is a very subjective question depending on who you ask. So, you know, for some teams that may well think that. Oh, is it the content does not work for my team that’s bad quality. And for some teams that would think that hey, whatever you translate has to be really faithful to the English source. So while I think a challenge for a big international company operating, having different branches, operating in different countries is that, you know, the content could be made here in the USA, and we’re translating correctly, according to our, you know, original source, but somehow that content just is not applicable when it goes to the regional offices. So, the regional office will say this is bad quality but actually it is good quality if you look at it it’s, you know, faithful to the English source. So a lot of times I think, imagine the localization program you kind of have to be that middle person to connect the city team with the headquarter team, and help them understand what the issue really is, sometimes the lies in the source. So, I think, quality is very different depending on how people see it. And for headquarters, a lot of times they want things to be consistent. So, they want to be translated consistently, you know, whether it’s in the region or the product and all the marketing materials. A lot of times you have to be that middle person to corral everyone together, tell them, hey, this suffers our translation memory if we translate it differently. You know that and this will suffer the SLA because if we have to constantly make tweaks, just to fit one side, then it will not work, so you’ve got to bring everyone together and have them agree on something, that’s, that’s definitely one of the challenges as well.

G: Yeah, cuz I mean you were mentioning in the beginning this whole thing about like quality, but also time, right, in this constant balance where, you know, it also you know coming in, depends on your budget how much money you have, and it’s like you can’t have all three of them, you know like quality, have it be affordable, and then fast. So, I’m just curious. So, with respect like the quality and time, are there other positions where you I mean I’m assuming that you have to sacrifice one over the other or, or how do you manage that situation, especially given the fact that you have the we’re talking about the constant localization and all that. How does that, like, pose a challenge as well.

M: Well one way Autodesk is trying to solve that challenge is to really advocate, our stakeholders to kind of employ and kind of turn on our machine. We have a machine translation, machine translation or Machine Learning Program. We really want them to start using machine translation, as the first step of translation and after that, we can have post editing with humans involved. Doing so it really lets our machine, learn the content that is specific to Autodesk, and so it can only get better and better. The more we use it the better it will get, right? And then with post editing, a pair of human eyes will be looking at these machine translated content and then kind of correcting the TM and then ensure that it will only get better in the future, so I think our middle ground is is really just to advocate, this new where we believe it will really help on the long term, deploying machine translation, and it will help the SLA while really increasing the quality.

G: I yeah I mean on paper that sounds, makes total sense right. I’m just curious of how is it was it is it is that an easy sell to the people that are not a part of the localization team internally? Right, because like I mean people kind of have a tendency, who aren’t exposed to like machine translation kind of just associate it with like Google Translate right, so what is the experience been trying to like evangelize MT within Autodesk?

M: That has been very hard, because, as you can imagine what people hear machine translation. Oh my god it’s like Google translation but yeah what we mean by machine translation really isn’t just machine translation is, it’s just turning on that pipeline for us to have the tool start you know recording, and then refining to our internal content with still having a post-editing by a real human. So, I think kind of the rhetoric has to change when we approach different teams, and really let them know that this, even though we call it machine translation. That’s just part of the process right like it’s not like everything is machine translation. And the second thing that I have been doing with my stakeholders is I would run tests for them. I tell them, “Hey, why don’t we turn off for only this language, and then for the rest we will still translate using full human, you know, procedure”, and then let them compare, let them have the regional office look at it and see if there’s if the quality is good or if they see actual, a lot of difference. So, these are kind of like the A B testing that I’m just throwing out to allow my stakeholders test. And you know what, with a with a purpose of letting them know that hey you can save on budget, and you know you as we do this, it can only get better, so it’s it’s a hard sell, but I think it definitely is where the ship, how localization should go forward.

G: Right and I mean I definitely agree with that especially like once people will kind of better get a better grasp of. That machine translation can improve over time when you start feeding it things that, you know, is only material related to what you do. Right. As you were mentioning before, I was curious because you said something interesting that you said that the rhetoric changes depending on the stakeholder you’re talking to. Can you elaborate on what you mean by that, when exactly do you mean by that?

M: Basically, in our field and localization field when we say machine translation we know what it does and we know that we use it as an aid, you know, to help our whole translation pipeline, but for stakeholders, they, they get stuck as soon as they hear machine translation, exactly what you were mentioning they think of Google translation, they just think of, you know, clicking a button and then that’s all they see, that’s not how it works.

G: So, not yet.

M: You just have to really prep them, and really tell them how the process works. A lot of the times instead of just saying, Oh, this is just, why don’t we just turn on machine translation, because I just had to say. There’s a few choices for you to pick from we can do have, you know, have a pair of eyes, translate first and and have someone else have another pair of eyes review it, or there’s another route where you can, you know, use this new technology, or we can use the machines to translate the content first. Keep in mind that the machine is specific to the machine translation is specific to content that’s related to your program. So, in the future, it can be better and better and have someone review these, the content that is translated. And then, of course, at that time, most people won’t buy into it because they think that the quality will still suffer. Right, so right. So that’s when you can really offer free testing or say like, why don’t we turn on based on our data we know that our, you know this, in this language and this locale, not a lot of people are you know viewing this content. Can we try using machine translation in this in this locale first, and you have the city team to get feedback. And we can you know we can translate that using machine translation and non-machine translation and do them both, and then have them, tell them that, you know, provide feedback, what’s better and what’s not. And we’ve done that for a program I’m working on and they do not see a few, a lot of a lot of difference. They couldn’t even tell.

G: Yeah, I was just going to mention that. They just compare them and they can’t even tell the difference. So I just, I just realized up until this point I kept calling it constant localization. Why didn’t you correct me??? It’s called continuous localization

M: I think people call it differently.

G: Oh, do they? I wasn’t sure. So I just want to go to the last question which is, so we were mentioning in the very beginning that the speed bump that you mentioned the challenge with continuous localization and then kind of this, this gap that it creates because there isn’t enough time to educate the vendors on the content that’s being translated, and so do you envision MT being a solution to this or how does that tie into this speed bump?

M: I don’t think MT will be the solution to fix that. It really still depends on your stakeholder relationship. And we unless you can develop a better tool you know like that, where we can have the stakeholders really invest the PRDs, or screenshots or something in that in the strings that are associated, but that we really need to, you will really need to have a good internationalization team that is working on these tools, so I think that’s the stage, in a perfect world that will be the best and your cycles of will have to be, you know, coding with the right references and that I think that’s very hard that’s a lot more, that’s a lot to ask to the engineers because I don’t think they would think that they were hired for to do that, so I think the most reasonable way to solve this is still evangelization based on you know trying to be proactive and then tend things where and let them know that localization really needs to happen at the design stage, and not the last stage when you’re done with everything and you’re just trying to push translation through. That’s, that’s really not how it should work, and then I think constant reach out and education and you really be being friends with the product managers and let them know that you need to be included at that very first stage. That’s the most important.

G: Yes, super interesting. Alright Melvin, I just want to say thank you so much for making time to come and discuss this speed bump challenge with me today. I just wanted to ask, do you have any last words for our listeners? Maybe some life advice? Or I know you love to go out to eat because I’m on your Instagram all the time, for places to go in San Francisco?

M: Right, um, I just recently joined this plan called meal plan where you can really, you know pre order food and then see which, among your coworkers are also going to get food from that restaurant and that has been pretty interesting. But other than that,

G: We’re not sponsored by meal plan just so everybody knows, like we’re not affiliated with them at all.

M: As far as for like life plans, I think, I think, really, like, I’m passionate about languages, I’m still really interested in working with people from different cultural backgrounds and I think being in the localization you really like you benefit from meeting different kinds of people and that has been very interesting, but I think what really helps you get through and really, you know, do well on the job is really your, your ma,  like project and project management skills and being able to negotiate with your stakeholders and your vendors. I think that has been the most, like most important lessons that I’ve learned from the roles in the localization field.

G: All right, that’s great Melvin. I look forward to being out there in the real localization field after I graduate and hopefully, well I know, our paths will cross again.

M: Thank you.