The Internet is a worldwide phenomenon and that has caused increasing headaches for application developers as more and more devices connect. Smartphones and tablets in particular have extended computing's power to huge swaths of people who never had the inclination or money to buy computers. In turn, that has created a challenge with localization -- that is, making apps sensitive to language and the nuances of language.
Just consider the differences in reading direction. Where many languages are read from left to right, others are read right to left. Then there's the issue with capitalization. According to Microsoft's Developer Network page, when using ASCII, simply adding or detracting 0x0020 to a letter's corresponding "code point," you can create uppercase and lowercase letters. It's not so for Latin numbers and letters with accents, meaning you can't just add or subtract a single value to and from characters to get the results you're looking for.
Other challenges include formatting, user interface, and string-related issues, as well as:
- Code pages that list character codes in a specific order have to use, in some cases, special identifiers to reference the code pages. In other cases, as with Chinese, Japanese, and Korean, the double-byte character sets needed won't allow the combinations of these languages.
- Complex scripts like Arabic, Hebrew, Thai, Vietnamese, and the Indic family require special considerations to get the right displays.
- Hard-coded font names and font sizes result in fonts not being displayed correctly, or being displayed illegibly.
- Input Method Editors that allow people who can't use standard 101-key keyboards (because they don't accommodate their languages) have to be supported in the application.
- Line and word break algorithms are different for Asian DBCS languages and Western languages.
- Function and short-key combinations have to be carefully thought out because of the differences in keyboards across the globe.
Using the crowd to translate
Underlying all of that is what many might consider to be the simplest aspect of making apps localized: language translation. The reality is, it's not all that simple, and the founders of Ackuna, initially a side project of Translation Cloud, continue to evolve the options available for translating."As the side project of Translation Cloud, Ackuna started off as a way for us to assign proofreading jobs to freelance translators," said Matthew Bramowicz, VP of operations at Ackuna. "So, say someone used Google Translate and it was a messy translation, they could have it perfected at Ackuna. As Ackuna gained a community we decided to switch gears and use Ackuna's existing framework to provide app developers with crowdsourced translation. The business model at this point is based off of Translation Cloud's professional services. We maintain Ackuna as an active translation community for free, but it also serves as a door to Translation Cloud. So, developers using Ackuna can also opt to have their projects professionally translated as an upgraded option. Also, the translators using Ackuna for practice can gain employment through Translation Cloud if they gain enough merit. The translation industry is competitive, and Ackuna is our unique way of gaining attention and nurturing discussion about translation and our services."
At its heart, Ackuna is currently a "gamified learning experience" for about 5,600 translators and 480 developers, but it's also a community where translators can network with each other and with potential clients. The translation process begins when a developer posts a project. The system is getting two to three new projects posted each day. Then, the crowd of translators go to work, with the turn-around time depending upon how much the developer promotes the project on social media, how completely the project was described and illustrated, and the number of languages selected.
Translations are free, with translators performing their craft for not only the experience and community aspects but also to take part in application development. Michael Duke, marketing intern at Ackuna, says translators are also inspired to translate apps they've heard about but can't use because they're not yet localized to their regions.
Addressing the accuracy questionTo ensure accuracy, Ackuna uses a voting system where translations must receive a given number of positive votes. Duke says that as the user base grows so does the accuracy and the speed of the translations. He says the mobile applications being translated are not extraordinarily text-heavy, and the social nature of the service helps to normalize the translations, even though there may be more variations across translations in a crowdsourced environment. Ackuna also encourages developers to provide as much context as they can in the form of screen shots and descriptions.
The service is considering making it possible for project posters to specify the geographic locations of their app's translators. However, Ackuna's social nature already allows developers to message specific translators if they happen to like their translation style, or even if they find out they are from specific regions that the app is aimed at.
Bramowicz emphasizes that it's best to understand what markets you want to get into before making a stab at localizing, as entering some regions is unlikely to be cost-effective or time-efficient. Currently, he says the romance languages such as French and Spanish are being translated quickly and that most apps arrive at the service in English. Chinese, Japanese, and Russian are other hot "translate to" languages.
The most difficult translations are those that come in unique language pairs. An example is a project translating Arabic to Chinese. While there may be large numbers of users on either side, there aren't that many translators.
More about the user base
So far, Ackuna is attracting mostly individual developers who don't necessarily have the budget for professional translation or who are exploring the prospect of localization. Bramowicz says that about 70% of projects are first-time projects and the remaining 30% are from developers posting multiple projects. While he admits it's possible that wholesalers could use the service to have multiple translations done that are then sold to others, he says it's unlikely, as developers could see their projects being uploaded to the service. He also says it wouldn't be very time-efficient, since social translation is slower and feeds off the developer's involvement. Those paying for translations would tend to expect them to be completed quickly.
Ackuna currently covers most mobile platforms and supports 10 file formats and 22 languages. Duke says jobs that don't get translated are often ones where the developers don't promote or don't provide enough context for the translators to translate accurately. Another reason translations don't get completed is when the developer decides to choose translations for every single language offered. In doing that, they may get three or four languages completely translated, but since they don't get all 22 languages completed, the project can't be counted as a total completion. Still, it's hard to say such jobs will never get completed, and for others it's rare they will take more than a few weeks.
Also read on TechRepublic: 10 tips and best practices for software localization
Duane Craig reports and writes on technology, construction, finance, food, and agriculture. He's been published in trade print magazines, the Washington Post, and widely on the web.