The Secret to Speedy Crowdsourced Translation

The Early Bird Gets the Worm

When it comes to crowdsourced translation, the sheer scale of the job can cause problems. The average app or website has so much content, and sometimes the community just can’t translate all of your materials very quickly.

Meanwhile, by the time you’ve translated every part of your product, your competitors may have already moved into that new market and established a foothold that’s hard to challenge. Fortunately, with a few tricks, you can learn to prioritize your content as effectively as possible, allowing for fast, accurate crowdsourced translations. You’ll be around the world in no time!

The Trick: Content Prioritization

You want a crowdsourced translation. You need it fast. Here’s the one phrase you need to know: content prioritization.

Simply put, you don’t need to wait until the completion of a translation project to launch your product in new language versions. It’s much quicker to identify the most important content, get it translated, and then deploy it immediately. The earlier you launch new language versions, the earlier you can test the potential of a new market or a non-English speaking community (for more on testing apps in new markets, see this post). An earlier launch helps you plan a more sophisticated marketing campaign. Plus, first comers in a market tend to enjoy more media exposure and other PR opportunities.

How to Prioritize Content Quickly and Accurately

You can prioritize your content manually—just look at the web portal and mark the content that you think is most important to translate. But is that content really important? Or is it just you who thinks it’s important? A data-driven approach is more reliable. That way, content is prioritized according to the users’ needs, not your gut feelings.

You can get useful data by harnessing user feedback, or through automation. Both methods offer simple and quick solutions.

Solution #1: Feedback and Voters

Always begin a crowdsourcing campaign with your community. Remember that the magic behind crowdsourcing is that the accumulation of small tasks can generate outstanding results. By letting your users choose the content that’s most important to them, you make their experiences central to your consideration.

Voting is a powerful tool for crowdsourcing decision-making. Set up a voting page for members of the community to decide which parts of the work require the most attention. If you prefer to have fewer people involved in content prioritization, you can consider asking for help from your community manager and some of your superusers.

Keep in mind that voting is a form of user contribution. Be sure to recognize this contribution.

Solution #2: Automation

Another solution to content prioritization is to let the data speak for themselves. You can use analytics tools—think Google Analytics—to automatically record traffic data, such as page views. The volume of page views indicates the importance of the content, and thus its priority for translation.

You can also use other web analytics metrics like Page Authority (PA), developed by Moz. PA shows how a given webpage is likely to rank in Google.com’s search results. It is calculated based on data from dozens of factors tracked by Moz, such as the number of backlinks. It can also indicate how important a given part of your content is in terms of organic search reach.

You can also implement traffic trackers for all strings of content that will need to be translated. The tracker is able to record the number of views by user, and it is useful to order the priority of the content at the string level.

Organizing Prioritized Content

After prioritizing content for translation, you need to show the content to your translators in an organized way.

Divide the Content into Chunks

The first step is to split the work into small pieces. The actual size of the chunks depends on the nature of your translating community. If your community consists of a few highly committed members, you can chop the content into chunks that can be finished within a couple days. If your community has a lot of users, you can divide the content into bits that are only one or two sentences long.

Display the Content Priority to Translators

You should then sort the content for your volunteer translators according to priority of translation. You can even display a priority bar for each bit of content. Still, the content priority should only suggest which content you want translated first. Make sure to let your crowdsourced translation team vote with their feet by working on the content that seems most important to them.

Case Study: Airbnb

Airbnb has used content prioritization to get a quick crowdsourced translation. Through prioritization, Airbnb managed to launch a website in a new language in under a week.

As a global travel rental company, Airbnb wanted to launch a natively translated Japanese website. And, because they have a passionate community of Airbnb hosts, they chose to use crowdsourcing.

However, Airbnb’s websites, email templates, and mobile apps contain 400,000 words of English content. That kind of translation can’t be done in just a couple days. As a result, content prioritization becomes essential for a speedy localization.

Airbnb set up a translation infrastructure that counted the frequency with which users saw a particular phrase. They employed a  t() method on their Rails server. For instance,

The instrument t() let translators know how to prioritize phrases for translation. If the phrase was seen by a user but had no translation yet, it was automatically recorded on a collection of all strings for translation. The phrase view was stored cumulatively in the server at a site-wide scale. This enabled Airbnb to prioritize the phrases based on their importance in terms of visibility.

Taken together, Airbnb showed their Japanese translators a view of all pages/emails on their site, and ordered the pages in terms of translation completeness in Japanese and importance overall.

(Airbnb’s translation platform)

Airbnb’s features have been developed over the past five years, but their localization team only wanted to get the newer, more important parts translated. Therefore, the ‘importance’ column represented the number of page views in the past 4 days only. That way, Airbnb just needed to instruct their translators to start at the top of the list of pages/emails and work all the way down.

After adopting these strategies, Airbnb was able to launch Airbnb.jp, their Japanese website, in only one week.

Last Takeaway: Study What Local Users Might Really Want

One last piece of advice: know the need of potential users in new markets. The preferences and concerns of users differ from places. For instance, Western users might pay more attention to privacy-related content than Asian users do. Because of these differences, studying your potential user base is an important part of content prioritization. It can help to segment your users by geographical location, and then researching any distinctive patterns.

Your Turn!

Have you ever tried one of these crowdsourced translation strategies? We love to hear what you have to say about your experiences with speedy solutions in crowdsourced translation! Just leave us a comment below.

Stay tuned for upcoming blog posts about the best strategies, tips, and practices for crowdsourced translation. We want to help you make your translation project a success!

Feature photo: Flickr

Reference: Airbnb Engineering Blogs, Wiki4us