Introduction
Translation crowdsourcing is a popular choice for many client-side companies and language service providers. It refers to the situation where a crowd going forward and collaboratively doing the translation work, often without direct financial motivations. It has been successfully implemented by many business and technology giants, such as Facebook, Twitter, Adobe and so on.
Although translation crowdsourcing has reduced the cost of hiring linguists and translators, it requires the cost of advertising, management, and technical infrastructure. Why did many companies still choose it? Firstly, it is a brilliant way of brand promotion. Secondly, when the volume of translation is massive or the deadline is urgent, the power of crowd can ensure the high-efficiency. However, the biggest headache for crowdsourcing translation lies in its inaccuracy. Most volunteers are not professional translators, and they don’t have the responsibility to provide high-quality work due to its non-profit nature. Thus, how to maintain quantity and quality control of crowdsourced translation is significant to its success.
There are a bunch of popular tools can be implemented into translation crowdsourcing. In terms of technology, there are Lingotek TMS, Ackuna, Amara and etc. In terms of terminology, there are Tricider, TermWiki and etc. Each tool has its own pros and cons. The decision maker also can customize its own tool based the unique features of their projects.
How to maintain quantity control?
1. Gamification: Add game-like elements to the whole system. Such as design and set up a rewarding system to volunteer translators, where they can win badges when they reach a certain amount of workload.
2. Machine Translation Application: Apply machine translation function to the translation process. When a project starts, the blanks can be automatically filled by machine translation results, and the translator only needs to revise it instead of translate from the stretch.
3. User-Friendly Interface: Design a user-friendly interface. When potential translators see the interface for the first time, they can quickly understand each process, and easily be attracted to start volunteering.
4. Positive Affirmation: Give affirmation and recognition to volunteer translators frequently, such as reminders of their contributions, virtual gifts and so on.
5. Process Reminders: Design and set up process reminders such as milestones to constantly remind translators of their workload.
6. Mobile Apps: Design a translation app so that translators can not only work on computers but also on phones or pads. Design off-line function so that they can download the project and translate it off-line. This can increase the potential working hours.
7. Subject Oriented: Classify the translation projects based on their subjects. Volunteer translators can choose the subjects they are most interested in before translation starts. The classification can reduce the potential subject barrier and maintain the passion of the translators.
8. Deadline Alarm: Set up deadline function for the translators. They can set the deadline for a certain project once it starts. The system can send deadline alarms automatically to the translators and remind them to finish their work beforehand.
How to maintain quality control?
1. Linguistic Preparation: Before the translation project starts, prepare relevant glossaries as volunteer translators’ reference and a style guide for them to follow.
2. Vetting System: According to the requirements of different projects, build up a vetting system to select qualified translators. The standards include their certification, educational background, prove of language skill, related professional experience and etc.
3. QA Automation: Design and set up a quality assurance auto-check function, and apply it to the translation process. When translators work on a project, the typos, grammatical mistakes and other issues can be automatically identified and corrected.
4. Segmentation: Design and set up a segmentation system, and logically segment the content of the project. the segmentation should obey two rules: 1. Don’t be too short and hurt translators’ understanding of context. 2. Don’t be too lengthy and demand excessive linguistic competence.
5. Shared TM System: Design a cloud-based translation memory which opens to all volunteering translators. For certain terms, the system can provide highly suggested translation results for the translators to choose. It can boost efficiency and accuracy at the same time.
6. Voting System: Design and set up a voting system for peer review. Translators from the same language pair can review each other’s translated segmentations and give votes to them. The segmentation that receives the most votes can stand out from other ones.
7. Hierarchy Establishment: Design several levels to differentiate volunteer translators according to the votes their translation received. The translators who scaled to a higher level can receive the bonus, such as certifications, free memberships and etc.
8. Professional Review: Hire professional reviewers from LSPs, academic institutes and etc to review the final projects. This step can be conducted after peer review.
9. Feedback: Design a feedback entrance that open to all volunteer translators. Anyone can provide suggestions to the community manager.
Final Project ｜ Translation Crowdsourcing for Taobao
By Jiyi Zeng & Tianlin Li
Why Taobao needs translation crowdsourcing?
We noticed that your website is providing service to other countries and regions, yet it is only available in Simplified Chinese and Traditional Chinese. The Google Translate plugin could be enabled while browsing the site, but it seems that it is not working so well, especially when it comes to the text embedded in pictures, which may constitute the majority of the information on some pages.
We understand that the reason you are reluctant to add other languages might have been that the target customers you have in mind are just Chinese-speaking people in other countries, and they may contribute to a large share of the sales figures. However, in recent years, we have seen your business booming around the world, along with the increasing frustration experienced by non-Chinese speakers who tried to use the site. If you are looking to further expand your business internationally and at the same time are also concerned about the potentially high cost of localizing your site into other languages, we have a solution for you.
Introducing a hybrid model that combines machine translation, community translation, and professional review. We recommend using a different method to translate the different content on your website, according to the content’s business value. How do we determine the value of the content? We look at the visibility of it and the potential profit brought by it. For example, the information on the product pages posted by sellers is highly perishable; therefore it is not recommended that professional translators be employed to do the job. The reason is twofold. The time it takes to finish the translation and the high cost associated with it would make it not feasible.
How are we going to implement this model?
After thoroughly analyzing the website, we believe that there are two major components that vary in value. For each component, we have the different integrated method.
UI: This includes the homepage, help page, setting page, shopping cart, etc. Undoubtedly they have the highest visibility and potential for profits. In this case, we will invite volunteers to translate the content and have professional translators review, edit, and confirm the translation. In the workbench where the volunteers translate the website into context, translation memories, glossaries, and machine translation are also available for reference.
Product Pages: We are going to offer two options for the sellers that are looking to translate their product pages into other languages. They could choose to utilize Google Translate plugin to translate and publish their pages. Alternatively, they could pay Taobao a fee to crowdsource the translation. The fee does not go to volunteers; it is only for the cost of management and for using Taobao’s crowdsourcing tool. It is also optional for sellers whether they would want professional translators to review and confirm the translation at an extra fee. Sellers will post the work through the tool and it will either be claimed by volunteers or assigned by the system. After the work is done, the system will notify the seller that the content is ready to be published. To avoid untranslated texts embedded in pictures, for either option, sellers may only upload a text file to the system and the final deliverable will be in the same format. Sellers will have to complete the DTP work (if any) themselves.
Currently, since Taobao is offering its international service in Korea, Malaysia, Australia, Singapore, New Zealand, Canada, United States, and Japan, we propose that the target languages be English, Korean, Japanese, and Malay.
Resources | Tool
| We plan to partner with Lingotek to build a customized crowdsourcing tool that integrates an in-context workbench and a TMS workbench that enables communication not only between volunteers and the localization team but also among different departments inside the localization team at Taobao. |
Team and Roles
| We will set up an independent localization team in your company. The localization project manager will deal with issues reported by regional managers and liaise with other departments. The regional manager will manage a local translator community. Linguistic and technical QA managers will maintain the translation quality through linguistic and technical assurance. |
Financial Support
| This localization team needs funds for the cost of management and daily maintenance. Once the project starts, Taobao may allocate some funds as the initial capital. Later, the fees paid by the sellers who need crowdsourcing will chip in. Once this project successfully expands to the global market and boosts the sales, the financial source can be generated from future profits. |
Operational Process
| 1. The localization team will start the first iteration of the localization project.
2. The team will solicit feedback from volunteers, professional translators, managers, other departments and customers to find out potential problems and improve the localization process. 3. After all the target languages are launched, the marketing department at Taobao will analyze changes in page views and sales volumes in different locales. 4. The data analysis reports will help the leaders of the localization team with decision-making regarding the allocation of resources in the future. |
Quantity and Quality Control Practices
In this particular case, we worked out the best practices for quantity and quality control in the process of conducting this crowdsourced translation project for Taobao:
Video Presentation