The tech world is experiencing a surge in translation innovation. The sector has been growing in relevance as persistent advancements in communication technology continue to shrink our world. Major tech powers like Google and Microsoft have increased their R&D spend on new translator technology, culminating in an eye catching demo of an instant English-Chinese translator at a recent Microsoft event.
The increased focus on translation is predicated by the prospective earnings that can be reaped from the provision of machine-based solutions. Research firm Common Sense Advisory has estimated that the global market of outsourced language services and technology will earn £21.1 billion in 2012. Furthermore, the proliferation of smartphones has expanded the consumer base to an even greater extent, as demonstrated by the Google Translate app's recorded 50 million plus installs.
Hence the importance of persistent development in speech translation, which according to Microsoft still offers a word error rate of 20-25 per cent. Current iterations of the technology incorporate algorithms that utilise ‘linguistic rules’ which are expressed via the common model of matching a source word (in the native speaker’s language), to a corresponding target word (in a chosen foreign language), with further processing via a phonological corpus – a database that consists of the grammatical rules and vocabulary of multiple languages - being required to place the words in appropriate contextual order.
This is where machine translation tends to come a cropper however, as trying to replicate ‘natural language understanding’ is difficult due to nuances inherent to human language (e.g. sarcasm is determined by tone and not by the choice of words).
That is not to say that the present offerings in the translation sector should be dismissed out of hand as they currently give the user decent approximations of translated texts and spoken word. Therefore I will use this overview to give a quick rundown of the more useful real-time translators currently available.
This is by far the most prolific mobile translation application, allowing for text based translation in 65 languages and also providing real-time speech translation for 17 languages (including English, Arabic, French and Dutch).
Its “conversation mode” was originally released in ALPHA last year and currently remains an Android specific feature. The speech to speech function is operated via a single handset, where the user is required to first pick both native and foreign languages then after tapping the microphone icon, speak the phrase to be translated. The application then speaks back the translated speech and allows the foreign speaker to reply in their own language by doing the same.
Translations, though basic are fairly accurate but much like most vocal recognition technology it requires slow deliberate speech to operate effectively, which negates any natural flow to the conversation.
The free app - which is available on Windows Phone, iOS and Android - also allows for image translation giving the user the ability to take pictures of text written in a foreign language - particularly useful with road signs.
This Canadian company provides a social networking solution to real time translation by utilising plugins for most instant chat and messaging platforms such as Windows Live, Google Talk and Facebook Chat. The user is only required to select the output language and then begin typing in the chat box as they normally would. This feature is also available in a free mobile app for both Android and iOS called Ortsbo 2GO.
Ortsbo can translate 53 global languages - including English, Chinese and Arabic. The company also provides a Twitter service that allows its users to place a widget that can pull tweets from a specific account or hash tag and provide instant translation.
MyLanguage’s Vocre application performs live translated video calling for iOS devices in 31 languages. The user dials out from within the app and then is required to hit a record button before they speak into the enabled mobile device and watch as their words are converted to text. After making sure that the text entry is correct the user must then click reply and the text is repeated in the chosen foreign language by a synthesised voice.
Much like Ortsbo and Google Translates mobile offerings the app gives face to face functionality.
Hanashite Hon'yaku is exclusive to Japan and more specifically subscribers to the country's largest network provider, NTT Docomo. Much like Google Translate the service facilitates the use of face to face translation through a single enabled mobile, what sets it apart however, is its ability to translate actual phone conversations in real-time.
It provides spoken translation after a short pause as well as providing a text transcript. The user is required to dial out via a provided smartphone application, which allows for calls to be placed to overseas, mobile or landline. The phone translations are being limited to Japanese, Korean, Chinese and English, furthermore, by utilising cloud computing the accuracy of the translations are not limited by a phones specification.
This Israeli start-up has foregone the app route in favour of a dial up service that provides live call translations on any phone. The service currently operates in 100 countries and can translate 15 languages and dialects.
This is the only service I've listed that requires no Internet connection or software install, the user must first dial an access number (found on the Lexifone site) prior to dialing the call recipient. Much like the previously listed speech to speech services, translation comes after a momentary pause as a tone signifies when to begin talking. The service allows the listener to review the translation in their own language before sending, thus ensuring the message is accurate.
The company has deals in place with telecommunication titans BT Group and Telefonica, providing Lexifone with Euro-wide coverage.
Those are just some of the translation apps currently on offer that provide mobile real-time translation. Yet though such innovation is commendable these apps still fail to facilitate free flowing conversation.
That is where the future of this technology lies and software giant Microsoft believes the use of the speaker’s actual voice in translation is central to this. At the aforementioned event held in China last month the computing behemoth demoed its instant English to Mandarin translator with the spoken translated output replicating the speaker’s voice and cadence.
French firm Alcatel - Lucent also hopes to launch a voice replicator called MyVoice as a complement to its landline based translator WeTalk, which is a more immediate sign of this technologies growth trajectory.
Thanks to breakthroughs in telecommunications, the world is becoming a smaller place each and every day. As a consequence multilingualism is fast becoming a necessity, and hopefully we'll one day be able to speak to anyone, in any country, in any language, all in real-time.