Year: 2018

In Translation Management System course, our group completed a project from Clarins, a world-known cosmetic company. In the process, we accumulated our professional experience in client education and communication. We played around with WorldServer and understood its role in different phases of a translation project. Overall, it is an impressive journey for us. Many thanks to my colleagues Isabella Sun, Weixin Mo, Wanli Zhou, Muge Zhang.

Let me walk you through the whole process of our project. First, here is the basic information of our project for Clarins.

Project Description | Isabella 

1. Client Info: Clarins Group, trading as Clarins, is a French luxury skincare, cosmetics, and perfume company, which manufactures products, usually through high-end department store counters and selected pharmacies.
2. Project Overview: The content we are translating for Clarins this time is Clarins Skin Spa Brochure. The skin spa program by Clarins has been established in over 40 countries. They hope us to build Translation Memory and Termbase for them as well as translate the updated brochure. It contains 879 words (including words on pictures). Starting from April 6, our group finished this project within two weeks.
3. Tools and Working Phases: To facilitate the whole process and make sure the deliveries’ quality. We used WorldServe to organize our work and generate quotes, and we used SDL Trados to build translation memories. Briefly, our work was divided into three phases: preparation, production, and finalization.

We drafted the business proposal for Clarins before the client meeting, click here to see our proposal: 

Proposal for Clarins 

We held our client meeting in April and summarized the reflections on it as followed:  

Lessons Learned from Client Meeting | Tianlin 

• Client Education: Since our client is not from language industry, they have very limited knowledge of our typical workflow. Good communication needs to be set up from the beginning to avoid misunderstandings.
• “Why should they care?”: Since our client is a for-profit company, they focus more on benefits instead of features. We should follow this logic and explain more about the benefits of our chosen resources and tools. They don’t need to know every step of our effort, but what consequences our effort can bring for them.
• Contact Person: We should include a contact person to guide our client through the whole process, answering their questions or else.

After the client meeting, we delegated the tasks according to our proposal. Each group member did a wonderful job and met some challenges as well. 

Here are some screenshots of the localized file (major deliverable): 

Here are our deliverables, including the source file, the localized file, termbase and translation memory tmx file, and pseudo translation file: 

Deliverables

Here are the lessons learned in terms of different steps: 

Pseudo-Translate | Wanli

1. Random Uppercases: Since the original source file is in PDF format, after converting it to a word document the text layout changed greatly, with upper cases showing up randomly in the text, which although didn’t affect Pseudo translate, would become a problem in translation.
2. The Pseudo translation sees a lot of text bleeding.
3. Prolonged text continued on a second page which doesn’t have corresponding pictures.

Termbase | Weixin

1. Because the source file is relatively short, we can have more considerations for term base establishment. Most of the terms in our term base are nouns because nouns are usually the foundation for understanding sentences.
2. We need to pay special attention to the encoding. The term base file needs to be saved as UTF 16LE so that it can be imported to World Server. And also, if we export the term base, the terms will become mojibakes if the encoding of the program that opens this file is not correct.

Translation Memory | Muge

The client didn’t provide us with any translation memory. But we offered to create one for them as an extra benefit. After we translated the source text into Chinese, we used SDL Trados Studio to align the segments and created a translation memory. The TM file is in tmx format.

Desktop Publishing | Tianlin

1. Picture List:  The document from our client includes many ad images. While extracting texts from images, I found it’s important to include screenshots for translators’ reference, so that they can have a good grasp about the context and come up with appropriate translations. (Lingotek can do a better job in terms of linguist’s reference)  
2. Style:  Because the document is for commercial purpose, the localized images need to be visually attractive to grab the audience’s attention. When English text is replaced by Chinese, certain typeface and type size need to be changed as well. Our team needs to inform clients about this change and get their approval first. Plus, consulting a professional advertisement graphic designer within the budget may be a good idea, in order to ensure the high quality of our service.

When the project is done and closed, we also had a post hoc meeting. 

Click here to see the slides

Notes: 

All mentioned material is created and written by all team members:

Isabella Sun, Weixin Mo, Wanli Zhou, Muge Zhang, Tianlin Li

Terminology management is an essential part of language service. No matter you are in the academia or in the workplace, having a unified term bank as an easy-to-reach reference can largely boost the quality and efficiency of your work. Terminology brings a positive influence to so many roles:

• Experts: They could write subject-oriented documents using unified terms, which creates the same tone.
• Translators: Terminology provides a knowledge base for their reference, facilitating the process of their translation.
• Clients: Terminology can ensure the documentation of certain product is accurate, avoiding the risk of liability issues. It reduces the time and cost in certain ways.
• End-users: Avoid their confusion when using products.

In my terminology management class, my group decided to provide terminology service for DJI, a leading drone company in China. We overcame obstacles in the management and technology field and created out a wonderful plan. Many thanks to my team members: Kayla Munoz, Isabella Sun, Brenda Hu.

Scenario 

Client representative: Department Liaison

“Our company is decent sized. We have a few hundred people spread across eight or nine departments. My role was created to liaise with software engineering, technical writing, and the parts department. None of them can get along. You see, each department is contributing content to big product launches worldwide. But they can’t agree on single terms to use for device components. The engineers are using abbreviated terms so that content meets character limitations of the GUIs. The folks over in parts have their own identifiers for those same components. Then, we have the technical writers who get so mad when they’re not consulted about the abbreviations or identifiers used.

These issues we have with inconsistencies are only compounded when we send content to translation. Terms don’t match across content products, and our customers are starting to get really confused. We need a way to link all of these terminology sets together…”

Case Analysis

Company Name: DJI

Company Website: dji.com

Basic Information: DJI (Da-Jiang Innovations Science and Technology Co., Ltd) is a Chinese technology company headquartered in Shenzhen, Guangdong. It’s known as a manufacturer of unmanned aerial vehicles (UAV), also known as drones for aerial photography and videography.

Organizational Needs: As a company that’s accounting for 85% of the global consumer drone market, DJI is decently sized and still expanding. The following problem is a lack of communication among different departments. The department liaison can be harmful to the company’s future development so DJI connected us and hoped we could find a solution.

Appraisal on Stakeholders’ Need: Our client has been fully aware of the major problem that exists in DJI. They pointed out that the inconsistencies in translation caused most of the issues. Based on this, we believe our client will be cooperative and willing to provide us with the necessary materials to build a terminology database.

Organizational Gaps and Inadequacies: The inconsistencies in translation content caused inefficiency in cross-departmental communication as well as consumers’ confusion.

Solution 

Based on the scenario and case analysis above, our team developed a terminology solution for DJI. Click here to see our pitch:

DJI Terminology Pitch

After presented our pitch to the class, we also got a chance to see the fantastic work from other teams, absorbing the wisdom from our peers as much as we can. Although we chose different scenarios and cases, we have a lot to learn from one another in terms of terminology service. Here is a list of their strategies collected by our professor Alaina Brantner. It’s definitely a good resource for a future career in terminology management:

Strategies from peers

Reference

The-benefits-of-Terminology-A-guide-for-language-lovers

Notes

All materials are written by our whole team:

Kayla Munoz, Isabella Sun, Brenda Hu, Tianlin Li

In the Advanced CAT class, we had a chance to try the second one: Statistical machine translation. We chose a field, built up a bilingual corpus from scratch and trained our own SMT engine. As a graduate student coming from a pure liberal art background, the whole process is sort of torturing, including data collecting and cleaning, system maintenance etc. However, I learned quite a lot from it. Few people except engineers in the translation industry can go under the theoretical surface of Machine Translation, swift his or her role from a “user” to a “developer” and understand its pros and cons from an internal perspective. Many thanks to my incredible team members, Rory Hou, Wei Wu, Melissa Zhang and Cheng Song.

• Choosing field

A first headache we faced is: which subject field shall we choose? Firstly, we need to find a field with tons of existing bilingual text that is easy to extract. Secondly, the text from this field should tend to be “hard”, for instance, field-wise terminology, unified writing style, etc. Text like poetry, soap opera lines is definitely off our list. Eventually, we agreed to do the patent translation.

We chose the waterfall workflow for our project.

Here is our proposal for the pilot project: SMT proposal

After the kick-off meeting with our “client” Adam Wooten, we start our project. 

• Data Collection, Cleaning, and Alignment

Then, we need to collect enough cleaned-up data to build our bilingual corpus. A well-functioned SMT engine needs an enormous corpus to be the underlying support. (eBay’s corpus for SMT training contains 1.5 million words). Due to time constraints, we could not reach the standard but we tried out best.

Data Source: We choose WIPO Patentscope, UN documents and PKU Law as our major source of bilingual data. Data Cleaning tools such as Okapi Olifant can be used to clean up untranslatable strings or other irrelevant parts.

Text Alignment: After we extract enough bilingual text from those two sources, we need to align the English text to the Chinese text. Due to the huge amount of data, many alignment tools we used to use such as YouAlign fail to work without a paid membership. Fortunately, we found a very handy tool: TMXmall, to help us align tons of documents and export them as tmx files in a short time.

Closure: The bilingual corpus is ready to go. We also cleaned up some monolingual text for later use.

• SMT Engine Training Kick-off

After the data preparation was finished, we got into the Microsoft Translator Hub, and created a new project.

We imported data from the bilingual corpus to training, tuning and testing sections, and started the first round of training. When the training is finished, the system will send notifications to each of us, and we can see an auto-generated BLEU score to indicate the linguistic quality of our training. Then, we started to change some data and conducted the second round. The BLEU score indicates the quality of our data for each round, and we change our imported data in a new round based on its fluctuation. Our goal is to let BLEU score go up.

• Quality Review 

After each round of training, we review the MT results and come up with a score based on our quality metrics (using MQM standards). We set the scale of the score from 0 to 1000 to measure translation quality.

We calculate the time and cost of our pilot project. Obviously, it’s different from our initial speculation. We updated our proposal based on those found in the pilot project.

Here is our Updated SMT proposal: Updated SMT proposal

Plus, we summarized all we learned from this project: Lessons Learned

In Localization Project Management class, I had a chance to build up my first language service company from scratch, with my awesome team members: Helen Jung, Chris Healy, Muge Zhang, Jiyi Zeng and Isabella Sun. It’s definitely one of my best experience in MIIS. We named our baby — GalaXync.  We’re here to be your bridge across linguistic divides, connecting you to an audience of millions around the world.

We chose WordPress to build up our official website. Here goes our mission and vision page. 

Our Mission –

GalaXync is a multi-language vendor that specializes in three East Asian languages: Chinese, Japanese, and Korean. Our executive team firmly stands behind the idea that no matter where one is from, we can synchronize any form of content to be relevant and enjoyable to natives of other parts of the world. Composed of expert linguists and project managers with years of combined translation, localization, and consulting experience from a multitude of industries, GalaXync is the best partner to launch your company into today’s ever-competitive, ever-changing global market.

Our Vision – 

GalaXync strives to be the bridge connecting everyone in different parts of the galaxy and close the linguistic or cultural gaps between them. While our focus is on East Asia, we hope to continuously grow our services, customer base, and areas of expertise, becoming the world’s largest localization services provider with no language left behind. Contact us or visit us in our office, where our team is taking one step closer every day to our goal of bringing worlds together.

After rounds of discussion, we settled up a standardized translation and localization workflow goes as followed.

• Preparation

1. Inquiry: The first thing that happens in any project is that you reach out to us! One of our account managers will then meet with you to find out how we can help you best. What’s the scope of the project? What audience are we helping you reach? Do you have any budgetary or time constraints? With this information in hand, we’ll help your project go from impossible to possible and kick off a rewarding relationship for both of us.
2. File analysis: The next thing we’ll need is the files you want to be translated. Our hardworking staff of engineers will analyze the files to create a word count and then prepare them for translation, which will make our translation efficient and keep your files safe.
3. Team assembly: The project manager will determine the best way to meet your translation goals, assemble a crack-team of our best translators, and create a timeline. Other companies might wait until after the contract is signed to assemble a team, but we will never tell you we can take on a project before we know we have the exact right people to do it. With a full-professional, in-house group, we can assure the quality of our translations. Plus, if you know who’s working on your product, you can rest easy knowing it’s in good hands.
4. Initiation: The engineers and the project manager send their details to the account manager, who uses this information to compile a quote ( Click here to see a quote sample of a given project: Quote Sample ). The account manager then sends this quote to you, including a detailed breakdown of the various costs that will depend on the nature of the project, its size, and its speed. If you like what you see, we’ll write up a contract and sign it with you.

• Translation

1. Getting the project to the team: Congratulations! The localization of your product is now officially underway! The project manager distributes the files for translation, a translation memory (a fancy tool that keeps us consistent and efficient), glossaries, style sheets, and any extra references you provide to us.
2. Translation: Our team of translators works together to translate the text into the language(s) you need (be it Chinese, Japanese, Korean, or English) and deliver the completed translation back to the project manager right on time!
3. Editing & proofreading: The project manager will get the translations to our editors, who will compare the translations with the original text to make sure that all of your content is retransmitted faithfully. Additionally, our proofreaders will conduct a second round of editing, this time to make sure that the translations are good quality, effective text in the target language.
4. Re-assembling the files: Once everything is up to par, our engineers reverse the process they performed in the beginning, putting the new translations back into your full product.

• Review & Delivery

1. Quality assurance: Once the files are ready, we will make sure that the final product works, testing for bugs and triple-checking that the translations are effective in context. If you agree, we will invite third-party reviewers to review the final product. Of course, you will also give us your input!
2. Delivery: Once everything is just the way you need it, we will deliver the freshly-localized product to you, ready to release to your international audience.
3. Invoice: At this point, you will confirm receipt and we’ll send an invoice to you, detailing the rates and costs indicated in the contract, along with any changes that were made along the way.
4. Feedback & evaluation: To continue improving and refining our localization services, we’ll request feedback from you, and debrief among ourselves to determine what was successful and what created problems. Hopefully, you enjoyed your experience with us enough that we have the pleasure of getting to do it all over again with you in the future!

We designed our file system to manage tons of files, click here to see how it looks like (Click here):  File System

To ensure the unrivaled quality of our service, we also worked out a vendor recruiting process, which includes three parts: Resource Evaluation, Language Evaluation, Onboarding and Welcoming (Click here): Recruiting Process

We used Basecamp to track our project and ensure good communication among one another. We used ProjectLibre to schedule our workflow as followed (Click here): Schedule

As a for-profit entity, it’s always vital to engage and convince your clients, making them buy your service. Here is a case pitch to show how we demonstrated ourselves to a client (Click here): GalaXync Pitch

The whole process is an engaging journey for me to walk through and understand how language industry works as a “professional” from a holistic viewpoint. And more importantly, how to be part of a team, effectively communicate with other team members, learn from others and collaborate with others. In the business world, many things need to be written down in a professional manner. I’m willing to go under the surface of project management and explore more.

Note: All mentioned documents are designed and written by our whole team: Helen Jung, Chris Healy, Muge Zhang, Jiyi Zeng, Tianlin Li and Isabella Sun