As part of a team project for my Translation Technology course at Middlebury Institute of International Studies at Monterey, I and a team of 3 localization students simulated a localization request from Berkeley-based publishing company Parallax Press. Here’s the run-down of our project, the successes, and the lessons learned.
The Project
Our team, Apples to Oranges Translation, conducted a localization of two book covers of world-renowned zen master Thich Nhat Hanh’s Meditation Essentials series, pictured below.
We localized the front and back cover into 4 languages: Korean, Turkish, Chinese, and Japanese. Our first order of business was coming up with a quote for the client, Parallax Press.
I extracted the words from the jpg images using two methods to test how well they could recognize the text.
- Adobe Acrobat Pro DC
- Google Keep extract text function
- Adobe Acrobat Pro was able to decipher what was text, but when I tried copying the text into a Word file, things got wonky. I’m sure that even if I put it into Trados for pseudotranslation, things would not have worked out well. It’s probably because the quality of the original image was poor.
2. Google Keep was much better at recognizing text from the image. While it did miss the [], misspell Earth, and make other mistakes, overall, this was about 95% ready for translation. It took me just 5 minutes to make the necessary corrections to the text to prepare it for pseudotranslation in the CAT tool.
If you’re interested in trying out Google Keep’s image text recognition function, here’s some simple steps on how to do that.
Like I mentioned earlier, given how poor the quality was on our image file, it made sense to first extract the text to a Word document, pseudotranslate it, then use Adobe Photoshop to bring the text back into the image. Thus, we came up with a game plan…
The Quote
Our Statement of Work for this client included a basic project description, timeline of work, a thorough project cost breakdown, list of resources used, a breakdown of the phases of work we will undergo, and finally the list of deliverables.
While it may seem like a lot of information in 2 pages, it’s necessary to include all these aspects into a quote so the client knows exactly what to expect from our team. The last thing we want is the client going back to us saying they want to change the source content or question our prices after signing off.
Clear communication starts with clear expectations from both parties.
While our Statement of Work is not too different from the standard Quotes available in the market, I believe a key contribution we made was deciding to use weighted words to calculate our translation rates. You’ll find the calculation of weighted words in the second page under Word Rate.
I go over weighted words more below.
The Deliverables
After completing pseudotranslation and finding no issues with the document in translation, we went ahead with our translation of the original source documents into our respective languages.
Using SDL Trados, I uploaded an existing translation memory we created for this project and a termbase to help with the translation. After translation, I uploaded the completed target file to our Project folder. Our final deliverables included the in-context translations, updated translation memory and termbase, and even a thorough style guide.
A team member completed the style guide, while two members completed the desktop publishing for all of the posters. I thank them for their contribution! The result turned out fantastic. Shall we take a look?
Here’s a closer look… I will share the Korean, which is the translation I worked on and for which I found a font. Again, another team member was the one who did the DTP. I’d say she did a great job!
The Lessons Learned
In any project, there will be rocky roads. The same was for Apples to Oranges Translation. While we were successful overall, we did face challenges. Challenges that I am grateful for, because we were able to learn from these challenges and come up with a better final product.
Some of the challenges were:
- Localizing barcode and prices
- We didn’t specify whether to translate the prices until translation was already done! Had we discussed this during the creation of the style guide, we would have saved time
- Language codes
- There was confusion over how to name the language codes (using the 2-letter ISO language code with the country code after [ie. JA-JP or KO-KR] or just the language code) We decided with the latter
- Final deliverables
- While we successfully organized our deliverables in neat folders, in a real project, it would make more sense to deliver project packages using Trados than to individually export the files
- Termbase and Translation Memory
- Some members had trouble converting the Trados .sdltx to a .tmx file. We solved the issue by using Glossary Converter to create the tmx file
- Manual calculation of the quote
- Calculating translation rates while factoring in discounts from fuzzy matches can get messy… It would have expanded our quote to 3 pages and most likely confused our client with all the numbers. To reduce this as much as possible, we calculated the translation rate using weighted words, which I will talk about below
As a result of our challenges, we learned a lot from this project. The most important lesson being:
Conduct in-depth internal meeting(s) after client approves proposal to determine translation rules, discuss potential risks, and align project objectives and implementation.
— Linh’s Lesson Learned
For that meeting, it’s also important to have an agenda on hand to be time-efficient. The more experience a LPM has in this project workflow, the more fleshed out their agenda will be. We created ours on the fly but I will take this experience to start the basis of my own internal meeting agenda.
A Case for using Weighted Words during Quoting
Finally, I’d like to talk about weighted words. Although this concept wasn’t taught during the Translation Technology course, it was a big game-changer for me when I realized how it could simplify the quoting process.
Weighted words is especially useful when:
- translating to many languages
- factoring in multiple discounts for fuzzy matches in the translation memory
- the client wants a simple quote breakdown
So, what exactly are weighted words and how do you use them?
Consider how translation rates are commonly calculated using Japanese as an example.
| Type | Charged at | Cost/word (JA) | # words | Cost/word (JA) |
|---|---|---|---|---|
| 100% or repetition | 20% | $0.04 | 50 | $2.00 |
| 95% – 99.9% | 30% | $0.06 | 16 | $0.96 |
| 85% – 94.9% | 60% | $0.12 | 9 | $1.08 |
| 0%-84.9% New words | 100% | $0.20 | 206 | $41.20 |
| Total | – | – | – | $45.24 |
So not too bad. But imagine…the cost per word is different depending on the language/demand/supply and many other factors. In this example, Japanese was $0.20 per word for new words. A more common language like Spanish might see a cost of $0.15 or even $0.10.
To get a full quote, we would have to recreate this price breakdown chart for EVERY SINGLE LANGUAGE.
An easier way of calculating costs is using weighted words. With weighted word counts(wwc), the discount is already applied to the word count so you just multiply the wwc with the language rate.
Essentially, instead of multiplying:
Add [% of rate] x [cost/word] x [word] for every discount rate and for every language, weighted words use this calculation:
Add [% of rate] x [word] for each discount rate to come up with a single number. This is the weighted word count.
| Type | Source Words | % of Full Rate | Weighted Words |
|---|---|---|---|
| 0% – 84% | 206 | 100% | 206 |
| 85% – 94% | 9 | 40% | 5 |
| 95% – 99% | 16 | 30% | 6 |
| 100%, 101%, Repetitions | 50 | 20% | 5 |
| Word Subtotal | 281 | 223 |
As you can see, it just takes one table to calculate the weighted word count. Next, you simply multiply that number by the cost/word for every language. Like so…
| Language | Cost/word | Weighted Word Count | Price |
|---|---|---|---|
| Simplified Chinese | ZH-CN | $0.18 | 223 | $40.10 |
| Japanese | JA | $0.20 | 223 | $44.56 |
| Korean | KO | $0.20 | 223 | $44.56 |
| Turkish | TR | $0.18 | 223 | $40.10 |
| Translation Subtotal | $169.33 |
As you can see, calculating with weighted words lessens the steps you have to take to calculate the translation rate. This lowers your chances of making calculation errors.
For more information, check out these articles on weighted word counts:
All about your project and weighted word counts
Thank you for reading through this project. I hope you enjoyed reading about my process and got something out of it. Have you used weighted word counts in your project quotes? Do you have a better way of conducting pre-flight on a jpg-based file?