What Makes A Good Localization Project Manager and Post-Editor

•      Was this content helpful?

Was this content helpful?

Unbabel analyzed the patterns of post-editors. Can we do the same for localization project managers? Are PMs in danger of being replaced by AI? And how does the personality change the game and why is nobody paying more attention to it? HR sucks… Listen NOW at 07:45​.

The second part is about Facebook’s patent to measure machine translation quality. It’s a great example of using crowd and users to judge the quality. The majority of the industry still relies on the opinions of a few individuals – wrong! We need to go deeper and be more contextual as the lord Gary V teaches. Opportunity detected…

This is episode #9 of my speaking practice, also known as the Localization Podcast 🙂 #localization​ and #translation​ news across social media delivered to you by the power of my voice.

Timestamps:
07:45​ – What makes a good post editor and why we should do the same analysis for localization project managers 39:51​ – How to measure translation quality through user engagement


Andrej Zito 

Hey, everyone, this is Andrej. with another episode of localization podcast, I think this will be Episode Number nine, if I’m not mistaken. So disclaimer for those who are listening to the podcast for the first time. This is kind of like my kind of like my solo practice, where I train speaking and talking about various localization topics that I discover on social media without a lot of preparation. So I just go over content that’s on Twitter and LinkedIn primarily, I try to pick up interesting posts, from different companies or different people. And then I try to share the most interesting ones with you. And usually, I try to provide some extra comment or I share my personal experience on the topic, if I have something to add. This episode is recorded from Manila, in Philippines, where I moved one week ago, it’s kind of like my extended vacation, I’ll be here for five weeks. So I wouldn’t do a lot of content from here.

Andrej Zito 

And my vacation officially started last Friday after I wrap up things with global me. So now I have two weeks of vacation, so I don’t have to do my usual job. So I can only focus on doing my content. And maybe I’ll try to squeeze in a little bit of like greeting nothing. Although I actually was doing a lot of nothing. Last week while I was still working remotely for for global meat. So instead of just being super lazy and doing nothing, I really want to kind of like experiment with my life, and it’s kind of like a test for me like, like, how would my life look like if I only did content creation. Because it’s something that I really like, it’s a very creative process, and I get to do it on my own. Nobody’s telling me like, how I should do it, what I should do. So I really like this.

Andrej Zito 

So I’m actually kind of like thinking like, if this test or if this experiment proves to be successful that then then maybe after some time, when, when I’m back in Vancouver, I may try to propose like a different set up, like a different job description for myself. So that I have more time to, to work on the content creation. And yeah, I forgot, I forgot one important thing for those of you who are listening. So this is the first time that I’m trying something new. And I was looking at my YouTube yesterday with my friend. And most of my recent videos are really just like a flip of the podcast, which is basically just like a static static image. And maybe I have like these little timers for the different topics, but otherwise, it’s not very engaging. And it’s not technically like good content for for YouTube.

Andrej Zito 

So what I’m trying this time is that I’m recording actually, in three different places, that places but three different trade different three different sources. So first thing that I’m doing, what I was always doing is that I’m recording my voice on my microphone. Second thing that I’m doing is that I’m recording on my desktop so that I can show the articles that I’m actually going through. And the third thing is that I’m also recording myself on my camera. So the content for YouTube will also have my face and it will also show the articles that I go through. So that’s the desk part. So maybe this will lead to better engagement on YouTube. Because I actually still consider YouTube to be my main platform that I want to focus on is just that I was just doing the podcast. And I didn’t do too many new videos, despite the fact that my first educational video is doing like really great like, like 80% of the traffic comes from search.

Andrej Zito 

People are still commanding, commenting, they’re still viewing the video, I think I will soon maybe get 200 views. And I have like, I think more than 10 hours, or maybe it’s like 800 minutes. So even more, more than 10 hours of watch time on that video, but I still haven’t done anything more. And I want to fix that this week. Anyway, I think that’s enough for the intro. So if you’re listening to this, and you actually want to see the process of recording and you want to look at me, when I’m doing the recording, then please head to YouTube and to my Youtube or just search, Andrej Zito or you can just I think you can just search the localization podcast and you will find something there. And yet you see a lot more, and you will still get the same audio output, then you would on the podcasting audio platforms. So I think I’m talking a lot already.

Andrej Zito 

So let’s start with the content for this week. I think I have probably five articles that I picked. And we’re going back to Slater because I have two articles from Slater. I have two posts from Twitter. I think it’s two posts. Let me check my notes and Oh, no. Oh, fuck What? Where are we? Yeah, we have two posts from Twitter. And we have one last article from linked in a BB. Okay. So let’s start with the first article. I’m gonna take note right here. And the first article from Slater is what makes a good post editor research examines activity patterns of over 300 linguists. How would I go and talk about this one. So this research comes from two guys from unbabel which is I think they’re, they call themselves AI company.

Andrej Zito 

They do a lot of empty. And this is research made by Antonio gooeys, who is a research scientist at unbabel. And Andre Martins, who is unbabel is head of research. And they published a paper and titled translator to fac understanding and representing human post editors. The paper points to prior research on the effectiveness of post editors looking at a number of topics among them, the relationship between pauses and cognitive effort, the use of novice versus professional post editors for research purposes, and the impact of post editor behavior such as planning a hat and mouse. This is keyboard use on overall performance. So the Purpose The purpose of this paper was to build on such work and find out whether it is possible to identify a specific post editor based on their actions.

Andrej Zito 

What a meaningful representations of post editors could be built, that would allow researchers to draw useful conclusions and ultimately, whether these repetitious representations could prove useful in predicting the time needed to post edit a document. Good job I’m just reading so far. The research has started from the premise that the combination of machines and humans for translation is effective and then they reference previous studies Showing that humans are more productive when post editing machine translation rather than translating from scratch. I think this is something that was already established as long as the quality of the machine translation is decent so that the translator doesn’t feel like they have to like completely delete the machine translation and basically translate from scratch because then it’s not helpful.

Andrej Zito 

But otherwise, definitely, post editing machine translations is better, faster, and cost effective for the customers. Moreover, they hypothesized gaining an understanding of how humans perform the task of post editing, and which methods are most effective can help make the human machine interaction in post editing even more successful. Okay, so the next paragraph about identifying good posts editors, the study relied on a data set of more than 66,000 source documents, and involved more than 300 post editors working from English into French and German. The source documents for translation work customer service, email messages sent to an babbles translation service. According to the researchers, the data set was the largest of the kind release to date, and the only one we are aware of with document level information.

Andrej Zito 

So okay, so customer service email messages. Okay. Oh, yeah, yeah, I think I think I checked the unbabel website. And for those of you who are looking for those of you who are listening to the audio, I’m actually going to show this on my video, YouTube. So unbabel is actually and I think it’s right here yet it says Alan babble, seamless, multilingual support. So I think their main focus is on translating, customer support, messages and content. So I’m wondering if the word is the article. So I’m wondering if the source documents that they say, is something whether this is like, customer support and messages that were sent by unbabel customers and users to them, so they could basically leverage their own data or if they used data of their customers. But judging from the amount of customers that they have, and the big names they have, maybe it’s the first one.

Andrej Zito 

I’m not sure. But I don’t even think it’s important to actually know where this data comes from. The data set was the largest, okay. So now this is important. This is this is interesting. The researchers look at common post editing operations, such as inserting, deleting, and replacing a word or block of words, and also took into account keystrokes, mouse actions and waiting times. From the way these operations or action sequences were carried out by individuals, they hope to identify specific post editors, and do so more reliably than they would simply by simply comparing machine translated text with post edited text. Yeah, so this is a very, very, very key information that I have never thought of. And it’s like about like, under identifying the patterns at which post editors work.

Andrej Zito 

So of course, like if you just compare like the, the input, which is the machine translation, versus the output, you can still like identify, like the changes that were made, and you can compare kind of the quality of the post editors, but you but it’s still kind of like a black box, you know, like you insert, machine translation and post edited, text comes out. But what these guys did is that they analyzed like keystrokes and mouse whereas it’s mouse actions and waiting times. So waiting times, probably First, like, the people are, I mean, the post editors are looking at the text and they’re trying to come up like they’re they’re basically thinking like when they’re thinking when they’re not taking an action. So it moves the black box into white box. Because you can analyze the way that they do the post editing.

Andrej Zito 

And so the most important thing for me when I was reading and this is what if we actually applied similar to call it like, pattern collection, or like the way people work, not just to post editors, but what it also applied to translators on their own or like, reviewers, although like, actually, translators are probably like shifting, mostly to post editors right now. Although, of course, there are cases like where you just translate from scratch. But besides like, the linguistic experts, to me, the biggest question and idea actually is, what if? What if we actually analyzed how project managers work? Um, or I mean, like, in general, like, like, how everybody works? Like, what are their patterns? Like, when do they think like, what apps they’re actively using? How are they typing, writing, like writing is like one of the most important parts part of your job.

Andrej Zito 

So it will be actually interesting, like how people write and compose their messages. And also like, when it’s like, Is it different, like, when to do it to the customers, when they do internally, when they just use slack? Or glip? And how they organize the time what are the patterns? So this is like, this is like, really? I’m wondering, like, if anybody actually did any research on that, and if not, how I could actually start like research like that. But it would be super, super, super, super interesting. Because like, I’ve been always, like, for a long time in work, like in various companies. I always considered myself to be like, one of the most efficient people. And, but I was never like, asked, or, like, encouraged, like, to make other people more efficient. And if I was actually asked to do that, I think I would just basically like kind of, like, try to, like Shadow them, like, beat with them and see how they work.

Andrej Zito 

And I’m thinking like, if we actually analyzed, or like, I don’t know, like, at first, like, collect the data, like, like, how people are, like winter typing, when they’re using their mouths, how many windows they have open, how they’re switching between the windows. So what I want to say is that project management is not a rocket science. Even without like, any pattern recognition or anything like that, I’m pretty sure that most of the things that we do could be somehow structured. Maybe it would be like a big tree blecha decision, how’s it called decision tree or something like that? Like, if you encountered this? Yes, no, then go here and then go there. You know what I mean? decision making tree, something like that. How’s it called? And wait, where am I going with this? Yeah, so it’s not a rocket science.

Andrej Zito 

So most of the tasks that a pm does, I think are first of all quite repetitive. Second of all, the decisions that the project manager does, in most cases could probably be put into a decision tree, or somehow pretty much automated. So then the biggest question is how the project managers operate? Like, let’s say 10 project managers come to the office in the morning? What is the first thing that they do? To the star with the emails? Do they start with the most complex tasks? most challenging tasks? How many meetings do they have? What are those meetings about? You know, like really like trying to deconstruct what project managers actually do. Because like, as a project manager, your main goal is to basically get the jobs done on time within budget and deliver the scope that was requested, where am I going with this? I have no idea totally lost.

Andrej Zito 

I totally, I’m totally lost, because I can’t organize my thoughts. I didn’t prepare any notes about this part. I just thought it I’m going to just like, freestyle talk about how I think the research about POS errors could be applied to project managers. And then the other thing, and I still haven’t even fully explained the first level, about analyzing how project managers actually operate. The second part I was thinking about like, because like, we have all these AI and everything that’s trying to be done by a machine wouldn’t actually, the work of project managers be a very good candidate to be replaced by machines, because of all the previous things that I said, it is repetitive depends, of course, depends on each individual customer that you serve, and the requests and kind of like the maturity of the on the operations that you have in place for a particular customer.

Andrej Zito 

So like when you’re starting out, there’s like a lot of poking around a lot of learning. So this is where the human comes, and they can set things up. But once like the process is like fine tuned. You should be only as a project manager dealing with exceptions that go beyond the standard workflow, beyond the ideal workflow, or not even like the ideal one, or the one that you have already experienced. And you can kind of like automate the different ifs and else’s. And even when it comes to dealing with new things, how would a machine with all the learning and the AI be able to deal with this? Like, what like a PM, if there’s like a new situation encountered? And the pm would be like, okay, machine, I would do this, because of this and this and this? So would the AI learn based off that? And with the AI? Would we would we be actually able to teach the AI to to learn how to deal with the exceptions.

Andrej Zito 

And then once the AI makes a decision, then evaluate the result of the decision. And based on that, whether it was like a good decision or a bad decision, further improve its learning. I think we’re well, well, that’s what machine learning is about. Right? I have absolutely like no idea. I’m just like, really just like thinking from the very limited amount of information that I have on this topic. But so what is it that I’m actually trying to say? I’m trying to say two things. Number one, like analyzing the patterns, like the research did for post editors, I think it’s something that should be done for project managers as well. Absolutely. Like there’s there should be like some high level pattern, it’s not even pattern. It’s more like a high level structure. Like every project manager probably knows, like how they approach things.

Andrej Zito 

And they do these kind of like, intentionally or what they think they’re doing. And maybe they could like describe like, this is how I work. It works for me because this and this and this. And then we should compare what they think how they work. We should compare that to the actual data that’s collected as they work. So again, what windows are they focused on? How many How are they typing, how many keystrokes they do, what is the action with their mouse, and try to try to identify like some patterns and maybe touch dose patterns to the results to kind of like so this this would be kind of like a client data and compare that to quality data kind of like like what do the customers think about this project manager?

Andrej Zito 

What do the your colleagues or your vendors or your freelancers think about his project manager, and maybe from all these data you could get or like some, again, just like the research did like maybe like some cohorts, like different ways have different project managers operate and maybe based on that we could come up with like, some interesting results. But also, maybe maybe it’s also tied to kind of like a personality of different project managers. Because like, maybe like the people who are more people oriented, maybe they prefer to have more meetings, and they solve things differently, like in a more personal way, then there are people who are maybe more introverted, and they prefer to use emails more. And the question is, like, is there? Is there a more efficient way to do this?

Andrej Zito 

Should everybody buddy be doing good, or what is like the right way? To operate as a project manager, when it comes to like your personality? I think that like the personality aspect is like so underutilized. Like, like, I don’t know, like most of the, and this is like, where, okay, we’re going into, like, totally different area here. HR sucks, you know, like, they just put up like skills, like, we want these people, we want a team player, we want a productive person, and blah, blah, blah, but what the fuck does it mean? Like, like, what do you like your personality? And, um, let’s say I’m referring to the key, what is it my bricks, personality? You know, those four, four letters? What if, what if, let’s say, let’s say, let’s say, you’re a project manager, you somehow you ended up being a project manager.

Andrej Zito 

But let’s say your personality, like you’re not very good at through all the stages, until it gets fully localized. But if you are more mature, and you have more automated solutions, and the team of the people that are assigned, they can actually do like a very good job, then the value of the pm goes down. Okay. Is there something else that I was thinking about? No. Pattern recognition. Second thing is personality. And the third thing is AI, will AI replaced project members? Should they be replaced? And what can I do about it to start it? Okay, yeah, I’m going to go back to the article so that we are done with this. So going back to post editing, blah, blah, blah. This was just like my own personal journey into what I think I could get out of this article. I really cannot talk to me, I’m sorry. He untapped sorts of information.

Andrej Zito 

 So that’s another paragraph of this article. Still, Slater and still talking about the research from two unbabel guys. So here we go. The key findings of the study were threefold. First, that action sequences can be used to perform accurate editor identification. Okay, I like it, I think the same could be applied to project managers. Second, that they can be used to learn human post editor vector representations that cluster to get a similar editors. I guess this is kind of like creating different cohorts and how people work and edit. Third and crucially, editor representations can be very effective for predicting human post editing time. So based on how you work, they can predict how much time you’re going to need to edit. marketing’s explained that being able to predict the time someone will take the post edit content can give useful in the context of matching linguists to a particular text type.

Andrej Zito 

More Moreover, according to the Martins It may also be used to inform customers about how long we expect the document to be translated. Well, that’s pretty much the same thing, right? Like how much time is going to take but this is like what I don’t understand. Because like how much time it’s going to take How does the tie to like the quality or whatever? Like typically, like what is the trade off like typically, like you would of course, always want someone who can translate Sorry, I’ll get post edit faster, as long as the otter results are as long as the other criteria are on the same level assuming so what is the trade of like, if they translate faster? Does it mean that the quality is like slightly less?

Andrej Zito 

Or like, does it mean that like, shorten, okay, okay, maybe maybe like a short time, like very fast translators should be translating something that’s very critical, like customer support messages, or I don’t know, what is like a fast like, I don’t know, customer generated content or something like that. For like marketing campaigns that need like that have like, longer time, you know, to launch globally, you wouldn’t need such a speed. Maybe that’s what they meant. Okay, so Moreover, they said this, we are currently looking at ways to use this information for human translation quality estimation, that’s predicting how good a translation is, before sending it to the customer. This will allow us to detect eventual translation mistakes and reassign the task to another human translator.

Andrej Zito 

Understanding post editing strategies also makes it possible to design our interfaces to better promote those behaviors, Martin said it, he said, One behavioral insight that surfaced during the study was that human posts editors who spend longer times reading before starting to type tend to type fast and to always edit left to right. By contrast, those who type immediately tend to spend time jumping back and forth. So I think this kind of makes sense. Because like the people who think about it more, aka spend longer times reading before starting to type, they probably think about, like how the whole sentence should be translated, and what are the changes that they need to make. So they’re going to go from left to right, because that’s how the sentence flows, right?

Andrej Zito 

By contrast, those would have immediately they tend to spend more time jumping back and forth. So they probably pick like smaller things, and they change one thing they’re like in like, let’s say, first part of the sentence, then they go to second part, and they may be third part, they change something there. And they realize that maybe Oh, now they need to fix something there. So it’s more like to the first people who who spend more time before they start typing. To me, it’s kind of like it’s more closer to the, to the traditional translating from scratch, because they probably just posted it the whole sentence in one go, while the other people, they just make small changes here and there. So I’m wondering like, which one is better and which one would be preferred, like in which case, they concluded that the post editing process is a rich and untapped source of information.

Andrej Zito 

And it is the researchers hope that the data set we released can foster further research in this area. Yep. Okay. I finished the first article. Yes, finally, it’s 46 minutes. I was for a long time brothers. So that was the first article with a lot of my ideas on how this could be extended to project managers. Let’s go to article number two. Still from Slater, this is about Facebook. And Facebook recently patented its process for measuring machine translation quality. The patent relates to Facebook’s meta merit method of gathering user engagement data on machine translation output and using this data to improve the quality of its machine translations. translations appear in use feed in a banner or in an ad for example, they may be automatically displayed or hidden until the user requests translation.

Andrej Zito 

The idea hinges on the premise that the more people like that the more people like share and comment on machine translated post, the better the translation is assessed to be. If you then add weighting, weightings. weightings and normalize the results. user engagement can serve as a proxy by which to evaluate machine translation Now, this is just another great example of using the crowd to judge what is better. And instead of just asking the people, is this transition better or no, they actually compare the actual results of people interacting with the content that was machine translated. So it’s not so obvious. And it’s even better because it’s like more subtle, like you don’t know, like which output you’re being served.

Andrej Zito 

So it’s kind of like a be testing, right? You don’t know what you serve, but in the back there, collecting all the data, and seeing which content gets more likes, and stuff like that. So this is like a, like, the quality shouldn’t be judged, but like, by only few people. And that’s still kind of like, the current trend, but it’s the current standards, that you have a translation then you have a reviewer, and one person kind of like judges the quality. And then maybe someone like from the local offices who speak the language, they look at it, and they say it’s okay. And then it goes to the public, where it gets viewed by I don’t know, hundreds 1000s 10 1000s of people. And it doesn’t mean that these few people represent actually the majority of the, of the audience. Anyway, so as I was saying, like, I saw this post from Gary Vee, which was about quality versus quantity.

Andrej Zito 

And, you know, like, he posts like, a lot of content, maybe like, I don’t know, he said, like, up to 100 pieces a day, across all different social media. So in this post, he was basically saying that you don’t know what quality is. And the only way to determine what quality is and what resonates with your audience is to do like, a lot of different content. And just see what sticks, you know, because like, and I know this, like, I kind of like realizes, like, when I do my own thing, or, or like even basically, when it comes to like talking to other people it’s like or how how you perceive like yourself, and what is quality for you, it doesn’t mean that it quality for other people as well. So, yeah, I think like the whole industry has this wrong or like needs to shift or needs to change and transition into a more experiment. mode, versus then just relying on few people who judge the quality,

Andrej Zito 

It should always be the end user and the market to judge what is quality and what is not quality. So that’s that is that. Hello, hello. So that is that? Yes. I’ll continue with the with the article. So Facebook set in the pattern that it displays different translations, to different groups of users called candidate translations, how one candidate translation performs relative to another tells Facebook, which one is preferred by users and therefore better. From here, Facebook can then tweak its models to ensure that the preferred translation is favored, making it more likely to be used in future. Facebook explains the iterative process as being repeatedly applied in order to create a feedback system in which multiple candidate translations are generated using a model. The translations are evaluated for user engagement, the model is modified to favor the translation having greater positive engagement.

Andrej Zito 

The updated model generates multiple candidate translations and the process repeats. So as I was reading this, it reminded me of genetic algorithms that I basic that I briefly experienced when I was trying to get on to automated trading systems, which are generated by genetic algorithms. And it’s basically a continuous iteration process. The other important thing is that Beyond beyond assessing which candidate translations are preferred, Facebook may also be able to tell which groups of people prefer which translations. This is a great point. For example, since Facebook often holds information about a user’s age, gender and nationality, it may calculate engagements courts on this basis, according to Facebook, different translations may be generated based on the language patterns of different demographic groups.

Andrej Zito 

And an appropriate translation may be provided based upon an identity of a user requesting the translation or target group identified in the translation request. So this is a very important this is a very important information. And it goes back to again Gary Vee, and his emphasis on context and relevance. So, again, if we have like a big market, maybe, maybe we even say like, for example, like in my, in our experience, like, let’s say, we have a huge let’s say, we have a new marketing content that is aimed for the gaming audience, gaming enthusiasts, but what what if like the gaming enthusiast, that’s like 10 year old, and he needs like his mother to buy him a new processor. The messaging for that person will be probably different than for someone who recently graduated and had his job plus, or versus third person who would be like in his 30s.

Andrej Zito 

He’s still a gaming enthusiast, but he has his own money. Right? Yes, absolutely. So then it comes toevaluating which translations are the two, the best engagement?And what if this metric or this KPI should be evaluated within each different cohort? Does it make sense? Yeah, it does. Because like, I don’t know, like, let’s say, translation. And let’s talk about like, if it’s a translation motor, or just stick to the traditional thing, where like, let’s say we have a post editor. So translation from post editor a, or from translation Model A, performs best gives the best results across across the whole market. But what if the translation from post editor be or from the translation model be performs best within a small, small cohort of the total market? But because it doesn’t do best on the global?

Andrej Zito 

What if it means that you’re basically giving up on a good quality, it’s just that you’re going to brote which again, is what Gary Vee says you have like a lot of vanilla content instead of going very specific and contextual to each user. So again, what are the ideas here? Is that we talked about like experimenting. But what if the experiment needs to go needs to be more contextual and more specific, instead of just saying that? A is better than me? What if a is better? for overall, overall, so evaluating the quality on a total model would be good for like cost effective solutions, because like you only want to have like, I don’t know, run like one test. Or you don’t want to segment your translations based on different cohorts. But then if you actually want to do it really well, and you wanna go really deep, then maybe you would need to do a lot more experimenting with thin, specific cohorts in the market. Right. Does it make sense?

Andrej Zito 

I think it does make sense. So this is another kind of like a localization strategy. ideas that I don’t see implemented anywhere. Okay, let’s try to finish the article. In some cases Facebook shows prompts to users asking them to say what a translation is usable or understandable. The emphasis on usability rather than ratings is intentional. The inventors found that asking a user to rate a translation often yields inconsistent results. Because a user may not know On what basis they should be reading the translation. By contrast, asking the user whether translation was usable or understandable produces more consistent and more useful results. With his method, Facebook is aiming to simplify the usually difficult and time consuming process of identifying which translations are favored. and communicating this information in a way that a machine translation system can can consistently apply.

Andrej Zito 

It is an alternative to the blue score, which the Facebook paid and points out has several problems. The final part is Slater reached out to prolific machine translation researcher Rico syndrich lecturer in machine learning at the University of Edinburg for his assessment of Facebook’s newly patented quality evaluation system. So this is an academic person. centric said that he was skeptical about using big data and user engagement to optimize empty social media platforms and search engines have managed to show users more relevant content by optimizing for user engagement. I understand why there’s interesting in using user engagement, also to optimize empty, big platforms get this data essentially for free, and it aligns with their business objectives.

Andrej Zito 

But I’m skeptical about this direction, he commented, explaining his position summary edit that I’m happy to believe that improving the translation quality will lead to higher user engagement on the platform. I’m less inclined to believe that optimizing user engagement directly will lead to better translation quality. Moreover, he said, with user engagement as its main objective, there is a risk that translation systems will learn to produce text that maximizes user engagement, while sacrificing translation accuracy. To give an example, when translating product descriptions in an online marketplace, manually using sales as the optimization criteria for an empty system could reward the system for embellishing product and misleading users, but then translating the description accurately.

Andrej Zito 

So yeah, this is this is a good point, I think. I have nothing more to add. I’m running out of batteries, and I’m running out of my voice. Anyway, so I’m going to stop the recording. Now. We’ll just do two pieces of articles from Slater because I talked a lot. And I don’t want to go further because I’m kind of like running out of energy. And so so this was episode nine of the localization podcast. Just a heads up. As I’m just recording this. I will do a little bit more promotion for this one, especially since I have this video. So yeah, watch out for it. And what day is it today? It’s Monday 11:30am here in Philippines. So Tomorrow is my usual day when I need to release a podcast. And then because I’m two weeks behind, I think I will need to do one episode where I just where I just do two weeks of content, but I’m actually considering because like episode seven and episode nine, I just talked about the two most interesting topics that I found and it was still one hour of talking.

Andrej Zito 

So I’m thinking that maybe I shouldn’t focus too much on The quantity, but rather just focus on the quality, which, again, is contradictive to what Gary Vee was suggesting, and maybe I should just do like some smaller episodes where I just talk about No, no, no, that’s a bad idea. The most important factor for me is that I pick articles that actually can spark some ideas that maybe I could put into practice with someone, hopefully, within our company. And where I can add a lot of extra thoughts that I have. I just don’t want to I don’t want to read the articles and not provide anything extra. Anyway, I’m just talking a lot of extra bullshit. So this will be it for episode nine. Thank you for listening. Thank you for watching, if you’re on YouTube, and I’ll get to the editing right after I do my lunch and I will talk to you next week. Bye bye.

We’re always creating new localization content

Make sure you don’t miss anything. Be the first to get our upcoming newsletter by joining the mailing list now. 

If you enjoyed that, you’ll love these…

Why hello there!

Enjoy 10% off

on your first course when you join our mailing list.

* All information collected will be used in accordance with our privacy policy. You may unsubscribe at any time.