Contact our team

Crowdsourced Translation: Does it Work?

September 6, 2016

The World Wide Web has made it crowd-sourcing imagepossible to communicate at all times, across all boundaries and on an unimaginable scale. The communication revolution lies at the root of trends that are changing how we go about our daily lives such as the sharing economy and crowdsourcing, crowdfunding, etc. But can crowdsourced translations prove as effective as traditional methods of translation?

The Origins of Crowdsourcing

Crowdsourcing is based on the premise that collaboration taps into a collective wisdom that is greater than the sum of its individual parts. As early as 2004 James Surowiecki eloquently argued that a diverse collection of independently deciding individuals with access to aggregated information will reach better decisions and predictions than could have been made by any single member of the group in his seminal book, The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies and Nations.

According to Crowdsourcing Week, “…crowdsourcing is the practice of engaging a ‘crowd’ or group for a common goal…Thanks to our growing connectivity, it is now easier than ever for individuals to collectively contribute — whether with ideas, time, expertise, or funds — to a project or cause…”

Crowdsourced Translation for Social Media Content

In general, crowdsourced translation falls into two main categories: translation by a “true” crowd, i.e., non-professional translators (such as Google Translate users correcting automatic translations) and translations by a “crowd” of professional translators working on crowdsourced translation service platforms such as TM-Town, Gengo, Smartling and others.[3]

The rise of social networks has created both a need and a platform for real-time translation of user-generated content in order to support communication on a global scale.

An interesting example of mobilizing a “true” crowd for social media translation is Facebook’s very recent introduction of a new feature that helps people post Facebook updates in multiple languages. The user can either write the translations from scratch or edit text that Facebook’s translation software generates from the original post. With more than one billion people using Facebook every day, this crowdsourced translation feature will provide Facebook with valuable new data to improve the ability of its translation software to convert slang and other informal language from one language to another (the holy grail of machine translation!)[1].

An example of mobilizing a crowd of translators to provide real-time translation of social media posts on a massive scale is Stepes’ recent launch of its Twitter translation API, which leverages the company’s mobile translation platform to translate and repost translated tweets to designated accounts. Customers sign up for the translation service via their Twitter accounts. The API extracts the tweets to be translated and sends them to Stepes’ network of ~50,000 translators who translate-on-demand anywhere and anytime from their smartphones. The cost to translate is ~10 cents per word and customers can set daily spending limits.[2]

Does Crowdsourced Translation Capture Collective Wisdom?

On the one hand, crowdsourced translation – whether by non-professionals or professionals – is quick, scalable and relatively inexpensive. But in the case of crowdsourced translation by non-professionals, who’s to say that their translations are good or that they won’t abuse the platform to teach the system false translations? Thus, for example, Facebook itself has announced that it doesn’t automatically start using new translations picked up from user data. It relies on professional translators to verify that the crowdsourced translations are correct. [1] There is little doubt that crowdsourced translations are far more vulnerable to mistakes and mistranslations than professional translation, and could prove risky if used in a professional or business setting. [3] On the other hand, a hybrid process like the one implemented by Facebook where professional translators oversee the crowdsourcing process and verify that the resulting translations are accurate could prove effective.

So does crowdsourced translation work? The truth lies somewhere in the middle. Indiscriminate, uncontrolled crowdsourced translation is unlikely to produce quality translations in a reliable and consistent manner. But platforms that encourage professional translators to share knowledge and collaborate can most certainly create collective wisdom that will drive professional translation to ever higher levels. And crowdsourced translations working in tandem with professional translators can potentially provide decent quality faster and at a lower cost that professional translation alone.

References

[1] Tim Simonite, New Translation Tool Will Help Facebook Master International Slang, MIT Technology Review, July 1, 2016
[2] Eileen Brown, Stepes uses human translation around the clock to translate Tweets, ZDNet, June 14, 2016
[3] Pieter Beens, The dangers of crowdsourcing translations, June 15, 2016