February 12, 2014 at 17:36The "non-creative" side of game localization. How do we do that
(New entry from Marina Ilinykh virtualtomato , senior project manager at All Correct Localization )
In the comments to one of our articles there was a question about what, in fact, the localizer does with files that need to be translated. To admit, this part of our work seems to me the least exciting, but if at least one reader is interested, we will be happy to talk about this with the example of software that our team uses. Since the conversation is not short, we decided to divide the story into two parts: preparing and managing the project (functional for the manager) and working with text (functional for the translator and editor).
Illustration from Dageron.com
I think it’s not a secret for anyone that any professional has rather strong feelings in relation to the tools with which he has to work. If the tool helps poorly or even interferes, then the worker experiences righteous anger. If, however, one cannot refuse this tool, then the expression of righteous anger can sometimes go beyond the limits acceptable in a decent society. But we were lucky - our main working tool, the MemoQ program, if it sometimes leads us to the use of profanity, it is solely because of the complete enthusiasm for its functionality.
MemoQ belongs to the category of TM-programs. TM is translation memory. In my opinion, translating well without using such programs is possible only in a number of limited cases. And now we will talk about the part of the work that the project manager performs in MemoQ.
It all starts with the appearance of a file for translation. It can be Excel, Word, txt, xml, po, json ... Yes, anything. All operations with the file: statistics calculation, translation, editing, quality control - are carried out in MemoQ. After downloading the file from the program, the source text in it is replaced by a translation.
A feature of TM programs is that they segment the text (divide it into cells) according to one of the predefined rules. And if there are repeating segments in the text, the program will automatically substitute your translation into all these segments later in the text. TM - translation memory is connected to each project. It stores all the translations that you previously performed on this project. That is, you will never have a case when you have to translate a particular text twice. For this, TM programs are everyone's love.
For example, this well-known poem is replete with repetitions:
But here is how the translation of one segment is substituted into all the others:
Statistics show that in this text 67% of repeating segments. Instead of 246 words, you need to actually translate only 81 words.
On large projects (games with a volume of over 20,000 words), such a number of repetitions is a rarity. Usually their share fluctuates at the level of 5-18%. For example, in Rayman Legends, out of 21,000 words, 2057 were duplicates. And in one of my last projects on localization of the social game, the proportion of repetitions in the text reached 18.2%. Imagine saving on such a project if the game is localized, say, in 6 languages. So ask the localizer for a rebate discount!
As I said, the text is segmented according to predefined rules. The choice of segmentation rules is one of the first important decisions that a manager must make. In most cases, the text to be translated comes in Excel format. And then we usually choose cell segmentation. That is, one cell of the Excel table corresponds to one segment in MemoQ. Also quite often we segment the text by sentences. But there are completely unique projects where you have to apply something very specific.
Once we translated a project from Korean into Russian. The client had a certain budget, therefore, to help him save money, we made the segmentation “smaller” so that the project had more repeating segments. Without going into too much detail, I’ll say that there are really more repetitions, but we just didn’t take into account that the structure of the Korean language is very different from, say, Russian. And as a result, we ourselves, without knowing it, activated the Yoda regime. Here are the funny stories we got (each line was a separate segment in MemoQ):
Kesivion, quarreling with her husband,
Prepare a dish of reconciliation
Going.
Or here:
At the night market out of curiosity
Elena played in a casino and lost.
And my favorite:
Sea elephants are definitely crazy.
Previously, a lot of people came
to look at them , the tourist industry
blossomed. And for some time now
they decided in Superman t-shirts.
Attacking tourists.
Where's it going ?!
In the vast majority of games, developers use variables. MemoQ can make working with them easier. For example, do you have such a text for translation:
Get {X} [cash] for {Y} [feez]?
Do you want to refill for {X} [feez]?
Loot for {0} Fee'z?
The manager (and the translator) can (and will certainly) make a template for your texts according to which all variables and tags will be identified and displayed as uneditable segments (red flags).
After the implementation of such a simple procedure, tags and variables are not taken into account in the estimate for translation (a little more savings!). An indisputable advantage of this approach is the ease of substituting variables into translation. They are inserted by pressing a single key, and not by endless clapping on Ctll + C, Ctrl + V or typing them from scratch. This reduces the risk of error and reduces labor costs.
Very often we are sent texts in xml format with the question of whether we can translate this way. Of course, if the translator opens such a file in notepad, then he will see something like this: You
can’t translate “like that”. This document still looks pretty good. There are times when you simply get lost in the chaos of the code that you see when you open the document. The risk of missing something or damaging the code is too great.
The most standard MemoQ xml filter in less than a minute will bring the document into such a beautiful form:
At the same time, the translator works only with the text, but immediately sees the results of its work in the preview window. For more complex documents, you can create unique filters that are stored on the server and can come to the rescue at any time.
It so happened that a team of translators should work on the game. That is more than one translator into one language. How to split the text? Only some manager-esculap will take up the file itself. MemoQ can divide a file into exactly as many pieces as needed (I once divided a file into 12 parts). At the same time, the file itself remains untouched, virtual copies are created in which the translator can only make changes to the part to which the manager has granted him access. But he sees the whole document and the work progress of his colleague, as they work in one server project. To make things even clearer: we don’t exchange files with translators. They gain access to projects on the server and work only in MemoQ.
In such simple, and sometimes, of course, tricky ways, we prepare the project for localization: we calculate statistics for translation, work with tags and variables, distribute files between translators. Now you can not be afraid that your files mercilessly split up, ruin the code, and you also have to pay for the "translation" of tags. No normal localization service provider will allow this. We will soon write an article about what the translator and editor see in MemoQ, what advantages the program gives them, and how this affects the quality of localization. And this article promises to be great!
Part 1
In the comments to one of our articles there was a question about what, in fact, the localizer does with files that need to be translated. To admit, this part of our work seems to me the least exciting, but if at least one reader is interested, we will be happy to talk about this with the example of software that our team uses. Since the conversation is not short, we decided to divide the story into two parts: preparing and managing the project (functional for the manager) and working with text (functional for the translator and editor).
Illustration from Dageron.com
I think it’s not a secret for anyone that any professional has rather strong feelings in relation to the tools with which he has to work. If the tool helps poorly or even interferes, then the worker experiences righteous anger. If, however, one cannot refuse this tool, then the expression of righteous anger can sometimes go beyond the limits acceptable in a decent society. But we were lucky - our main working tool, the MemoQ program, if it sometimes leads us to the use of profanity, it is solely because of the complete enthusiasm for its functionality.
MemoQ belongs to the category of TM-programs. TM is translation memory. In my opinion, translating well without using such programs is possible only in a number of limited cases. And now we will talk about the part of the work that the project manager performs in MemoQ.
Download and statistics
It all starts with the appearance of a file for translation. It can be Excel, Word, txt, xml, po, json ... Yes, anything. All operations with the file: statistics calculation, translation, editing, quality control - are carried out in MemoQ. After downloading the file from the program, the source text in it is replaced by a translation.
A feature of TM programs is that they segment the text (divide it into cells) according to one of the predefined rules. And if there are repeating segments in the text, the program will automatically substitute your translation into all these segments later in the text. TM - translation memory is connected to each project. It stores all the translations that you previously performed on this project. That is, you will never have a case when you have to translate a particular text twice. For this, TM programs are everyone's love.
For example, this well-known poem is replete with repetitions:
But here is how the translation of one segment is substituted into all the others:
Statistics show that in this text 67% of repeating segments. Instead of 246 words, you need to actually translate only 81 words.
On large projects (games with a volume of over 20,000 words), such a number of repetitions is a rarity. Usually their share fluctuates at the level of 5-18%. For example, in Rayman Legends, out of 21,000 words, 2057 were duplicates. And in one of my last projects on localization of the social game, the proportion of repetitions in the text reached 18.2%. Imagine saving on such a project if the game is localized, say, in 6 languages. So ask the localizer for a rebate discount!
As I said, the text is segmented according to predefined rules. The choice of segmentation rules is one of the first important decisions that a manager must make. In most cases, the text to be translated comes in Excel format. And then we usually choose cell segmentation. That is, one cell of the Excel table corresponds to one segment in MemoQ. Also quite often we segment the text by sentences. But there are completely unique projects where you have to apply something very specific.
Once we translated a project from Korean into Russian. The client had a certain budget, therefore, to help him save money, we made the segmentation “smaller” so that the project had more repeating segments. Without going into too much detail, I’ll say that there are really more repetitions, but we just didn’t take into account that the structure of the Korean language is very different from, say, Russian. And as a result, we ourselves, without knowing it, activated the Yoda regime. Here are the funny stories we got (each line was a separate segment in MemoQ):
Kesivion, quarreling with her husband,
Prepare a dish of reconciliation
Going.
Or here:
At the night market out of curiosity
Elena played in a casino and lost.
And my favorite:
Sea elephants are definitely crazy.
Previously, a lot of people came
to look at them , the tourist industry
blossomed. And for some time now
they decided in Superman t-shirts.
Attacking tourists.
Where's it going ?!
Work with variables
In the vast majority of games, developers use variables. MemoQ can make working with them easier. For example, do you have such a text for translation:
Get {X} [cash] for {Y} [feez]?
Do you want to refill for {X} [feez]?
Loot for {0} Fee'z?
The manager (and the translator) can (and will certainly) make a template for your texts according to which all variables and tags will be identified and displayed as uneditable segments (red flags).
After the implementation of such a simple procedure, tags and variables are not taken into account in the estimate for translation (a little more savings!). An indisputable advantage of this approach is the ease of substituting variables into translation. They are inserted by pressing a single key, and not by endless clapping on Ctll + C, Ctrl + V or typing them from scratch. This reduces the risk of error and reduces labor costs.
Complex formats
Very often we are sent texts in xml format with the question of whether we can translate this way. Of course, if the translator opens such a file in notepad, then he will see something like this: You
can’t translate “like that”. This document still looks pretty good. There are times when you simply get lost in the chaos of the code that you see when you open the document. The risk of missing something or damaging the code is too great.
The most standard MemoQ xml filter in less than a minute will bring the document into such a beautiful form:
At the same time, the translator works only with the text, but immediately sees the results of its work in the preview window. For more complex documents, you can create unique filters that are stored on the server and can come to the rescue at any time.
Large files
It so happened that a team of translators should work on the game. That is more than one translator into one language. How to split the text? Only some manager-esculap will take up the file itself. MemoQ can divide a file into exactly as many pieces as needed (I once divided a file into 12 parts). At the same time, the file itself remains untouched, virtual copies are created in which the translator can only make changes to the part to which the manager has granted him access. But he sees the whole document and the work progress of his colleague, as they work in one server project. To make things even clearer: we don’t exchange files with translators. They gain access to projects on the server and work only in MemoQ.
In such simple, and sometimes, of course, tricky ways, we prepare the project for localization: we calculate statistics for translation, work with tags and variables, distribute files between translators. Now you can not be afraid that your files mercilessly split up, ruin the code, and you also have to pay for the "translation" of tags. No normal localization service provider will allow this. We will soon write an article about what the translator and editor see in MemoQ, what advantages the program gives them, and how this affects the quality of localization. And this article promises to be great!