New technologies offer a solution for people who do not have ready access to linguistic databases. Last year, three start-ups introduced products that translates spoken sentences directly. Consumers and business travelers can thus communicate with other citizens of the world without having to speak a different language. There is great interest in the products, but there are also pitfalls.

For many people a world without language barriers is an attractive idea. That explains why since the 1970s companies have been developing technologies and systems that can translate spoken sentences directly. In particular, the introduction of translation computers in feature films and science fiction books has fuelled interest in these systems. With a translation computer, individuals and companies can continue to speak their own language and still communicate with people in another language. This has many advantages: no interpreter costs, no need to learn a foreign language again and literally boundless possibilities. Whoever is able to speak all languages has the world at his feet.
However, it was still decades before the first real translation computers were introduced. Only last year, the Chinese Waverly Labs announced the first initiative for a consumer translation computer, called Pilot, which fits into a handbag, and was funded through the crowd-funding website Indiegogo. Shortly afterwards, two more companies followed: the Japanese company Logbar with ili and the Dutch-Chinese company Travis with Travis the Translator.

Instant translation

The remarkable thing about the introduction of the three different products is that they were developed independently of each other. The three products all meet the same needs, but differ greatly. Yet they all have a common denominator: they all offer an instant translation. A small device containing a microphone and loudspeaker is used. Users select a source language (the language that they speak) and a target language (the language into which they need to be translated) and speak their sentence into the device, and then, within a few milliseconds, they can hear a translation . This is not exactly direct communication, but an interpreter is no longer needed.

Mathematical translation

Pilot, ili and Travis are actually speaking variants of Google Translate, Google’s written translation machine. The software in the small devices records words and quickly calculates the correct translation. It is played back by the speaker using a text-to-speech function.
Because of the infinite number of word combinations possible, developers choose a ‘calculation method’ for the devices. It searches for the correct translation of each word, and then combines all translations into a single sentence. This has the advantage that there is no need to create a database with hundreds of millions of possible sentences. The disadvantage is that the translation computers do need to have a strong processor to calculate translations quickly.

Context-dependent translations

The method for calculating translations offers technological advantages and makes it possible to develop relatively cheap ‘pocket translators’. At the same time, there is a possible danger. The databases with translations are often compiled on the basis of comparisons of millions of words. This means that at the moment only a literal translation can be provided, without taking into account any database errors or the context of a particular word. Anyone in France who tries to explain in English that their dressing has become loose should not be surprised if the pocket computer translates it as “sauce” (salad dressing).
The fact that the processor takes care of the translation’s pronunciation can also lead to confusing situations. In languages where the emphasis can make a big difference, instant translation can lead to raised eyebrows and strange looks.

Languages available

Such “teething troubles” are not affecting innovative products yet. The Pilot and Travis the Translator collected hundreds of thousands of dollars within a few days via crowd funding. Consumers are eager to try out the products. There is no objection to the fact that only a limited number of languages are currently available. In order to anticipate technological problems, they offer their services through an app or via the cloud. Pilot initially supports a limited number of languages and will deploy languages based on user demand. In addition to 20 offline languages, Travis offers 80 languages in the cloud, while ili offers users the ability to upload languages from a computer. In the future, companies will be able to offer improved language packages that will help users have fewer difficult situations in foreign lands. Travis even opts for artificial intelligence, so that the device can take off each phrase of speech.
Yet these intelligent wearables will not oust the traditional interpreter. Users need to choose their words carefully and have to wait for the translation after each sentence. So a fluent conversation in two languages is still a thing of the future.

This article was published in Reformatorisch Dagblad on July 12, 2017, showing to consumers the current state of translation wearables and their drawbacks. Unfortunately none of the parties is willing to share a product for review until now.