Will the New Tweet Translation Feature Render Accurate Results?
Twitter recently announced the inclusion of a tweet translation feature. Realizing its role in relaying world news, Twitter enabled this feature on an experimental basis for a few high-profile Egyptian users in the wake of recent unrest there.
Having a translation feature is nothing new for Internet-based services where services like Google Translate already provide the on-demand translation for seventy-one languages. While Google's translation service is quite popular in this regard, Twitter is apparently making use of the Microsoft Bing Translate technology. Microsoft recently opened up Bing as a Platform for developers, and one of the capabilities is “Write Once, Read Anywhere,” which is powered by the Bing Translator Control.
Services like Google Translate and Bing Translate form part of a larger technology term called “machine translation.” Wikipedia defines machine translation as “a sub-field of computational linguistics that investigates the use of software to translate text or speech from one natural language to another.”
There are many approaches to facilitating machine translation, but both Google Translate and Bing Translate make use of Statistical Machine Translation (SMT). In SMT, the translations are generated on the basis of statistical models that make use of bilingual text corpus.
The accuracy of results from SMT is generally directly proportional to the volume of language pair data available. The jury is divided on whether Google Translate or Bing Translate is the better machine translation engine, but both seem to deliver comparable results considering some of the independent tests.
Facebook makes use of Bing Translate to enable machine translation for its pages. The social media giant's choosing not to use Google Translate APIs could arguably be attributed to Google's decision to make the Translate APIs paid based on the usage; Bing Translate is available for free.
A recent Stanford University researcher's work proved that machine translation providers like Google and Bing show a gender bias in translating certain texts—referring to doctors as men and teachers as women. Whether the tweets will be accurately translated using Bing Translate is always questionable, but the new feature will always provide a better idea of what the tweets mean in the native languages.
The language pairing support also has a bearing on the machine translation system. The accuracy level of English to Japanese may be different from the accuracy level of English to Spanish. Machine translation is a complex technology, and it can be safely assumed that it will not replace a human translator in the foreseeable future.
In the case of machine translation, sometimes the best output is the one that is least worst. Twitter seems to understand this well and has started small by supporting one language translation—thereby giving itself a chance to gradually improve the user experience.
