Imagine meeting a local while traveling and being able to have a full conversation with them solely via voice on a mobile translation app. This may soon come to reality. The National Institute of Information and Communications Technology (NICT), headquartered in Tokyo, has introduced VoiceTra. The app can translate text, listen, or speak in 16 different languages. It’s simultaneous interpretation function can detect sentence structures to appropriately start feeding translation while the other speaker is still speaking. NICT hopes to also develop image recognition to detect speaker’s facial translations as well. Imagine a computer being able to detect sarcasm! NICT flaunts a ~90% accuracy rate in certain subject matters and the more the app is used, the stronger and smarter it gets.

Google and Microsoft are now pouring their resources into developing the same type of software, if only out of pure technological jealousy. If either of these companies were to partner with other private entities, they could grow their data at an alarming rate. Being able to tap into other company’s translation databases would give a much larger pool to pull from for increased accuracy. If one this is certain, no one company will be able to pull this off alone.

But what about the accuracy? What potential problems could be caused by incorrect voice translation? Receiving the wrong meal at a restaurant due to poor translation would be frustrating. However, if you have a prominent food allergy, this could be quite fatal. And what about voice recognition capabilities? If you own a Google Home or Alexa device, you know how frustrating it is to try and get them to understand you, even when speaking clearly. Now suppose you’re in a loud store, restaurant or on a train, how will the ambient noise affect the voice recognition? Until these apps have 100% accuracy, keep in mind to use them in the right environments as a communication facilitator, not a replacement for human translation.