We now live in a world where, akin to a sci-fi movie, machines can translate between languages for us. The translation industry has undergone a profound transformation in recent years, with the advances in and accessibility of translation technology developing at a rapid rate. No longer do we have to rely on Google Translate.
From global mega corporations to smaller enterprises, businesses are collectively investing billions of pounds into developing progressive translation tech that is slowly but surely overhauling the complete reliance on human translators. Now, more and more innovations are informing how translation is carried out, and shaping the very future of the industry itself.
Neural machine translation is pushing the boundaries of AI translation
The development of Artificial Intelligence (AI) translation technology has breathed new life into the translation industry, providing faster and more accurate non-human translation than ever before. Some of the world’s largest companies, including Amazon, Microsoft and Google, have invested heavily to provide AI translation as part of their services. These companies have all turned towards neural machine translation (NMT), which has proven to be far more effective than its predecessor, statistical machine translation (SMT).
With SMT, predictive algorithms teach a computer to translate text. A translation model and a language model is created during the training process, with the former storing various translations of certain phrases, and the latter storing the probability of a sequence of phrases appearing. During translation, a decoder picks the best result from these two models, and whilst this is useful for translating phrases, it does not bode well for fluency and grammar.
On the other hand, NMT uses neural-network machine learning technology. Words and phrases appearing in similar contexts during the training phase are encoded into what is known as ‘word vectors’. This means that when translating between texts, NMT is searching for complete sentences, not just phrases. As the model is translating words, information and context, it is much more fluent than SMT.
However, NMT has still yet to reach parity with human translation. Whilst it has undoubtedly made substantial progress, some noticeable flaws remain. It still cannot translate as accurately as humans would, and can only translate on a sentence-by-sentence basis. This means that any documents it translates are essentially a sequence of translated sentences, rather than a coherent whole. Yet, with the natural language processing (NLP) sector’s value forecasted to rise from $7.63 billion in 2016 to more than $16 billion by 2021, AI translation should get closer to human translation as NMT technology develops. As Luigi Koechlin of Global Voices, who provide online translation services to global industries, states “the real value is now more on the quality of the data each player holds”, adding that data from the NMT process “is becoming a commodity, like gas or electricity.”
Mind Rockets’ translation technology is aiding the deaf
Translation tools for those who are deaf and hard of hearing are particularly lacking, with the majority of apps and services aimed at those without hearing issues. This is where Mind Rockets come in. The Jordan-based startup has developed apps that instantly translate text or speech into sign language, with avatars appearing on screen to sign these translations to users. Mind Rockets has developed two apps using the avatars specifically for sign language in different cultures: Mimix3D for American sign language, and Al Turjuman for Arabic sign language
Mind Rockets’ apps have received rave reviews from users, and in May they won the top prize at 2nd AIM Startup MENA Pitch event. Their innovation has enhanced communication between individuals and has proven particularly useful in enabling teachers to simultaneously teach both hearing and non-hearing students. This is just one of their range of translation tools, which also includes web and Facebook page interpreters—particularly valuable technology when you consider there is only a 20% rate of proficient literacy within the deaf community, which renders much of the internet inaccessible.
Real time wearable translators are making travelling easier than ever
Universal translators have also long been a common feature in sci-fi, like Star Trek, where a device was used to decipher and interpret alien languages, and there are now various products providing us with instant translation of languages on this planet. The ili device is aimed at travellers, and can verbally translate short spoken English phrases into Mandarin, Japanese, Korean, Thai, or French, and vice versa. Impressively, unlike other translation devices, ili does not require an internet connection, allaying travellers fears that they will use up their mobile data or require an extra SIM card whilst using their phones abroad. A similar invention called Transay—launched last year—provides two-way translation for up to 28 languages and dialects.
Whilst this technology is undoubtedly useful, it is still very much in its infancy; both of these devices have limited lexicons and will not help users have a full conversation or be able to cope with, say, complex directions. However, they do ultimately make the lives of travellers easier, and with further refinement they could be even more useful assets further down the line.
Although commentators are sceptical that machine translators will ever completely replace human translators, the recent developments in translation technology show that the industry is diversifying and providing more ways to access translation than ever before.