Artificial Interpreters
Artificial intelligence (AI) has been toyed with since the automatons of Hephaestus in Greek mythology, and in today’s modern age, artificial intelligence is no longer just an idea or for chess players. It is being used for transportation, finance, medicine, music, agriculture, and more. Of interest to us, is AI in the field of interpreting and translating.
AI is improving, but still isn’t reliable enough to bet your life on. A recent study from the University of California at San Francisco showed that Google Translate correctly translated 594 medical sentences from Spanish to English out of 647. This may not sound too bad until you read that 15 out of the 53 translation errors were determined to possibly cause serious injury. An even higher rate of potential harm was reported when translators read through the Chinese translation.
If you’re an employee in the medical field and are thinking that it’s good enough, you may want to think again. Your patients may not feel the same way if they realize that they have a 1 out of 50 chance of suffering significant harm due to a mistranslation. Especially since these errors are not as likely to happen if a professional translator had completed the task. Skimping on hiring a certified translator or interpreter may just end up costing you significantly more than any savings via AI translations.
What about Sign Language?
Sign language is much trickier for AI to recognize, but that’s not stopping people from trying to make it happen. Trying to help bridge the communication gap between the Deaf, who rely on Sign Language and hearing people is a noble cause. Some approaches use gloves as sensors that feed the movements into the program, but this can be uncomfortable for the user.
Michigan State University’s approach replaces the gloves with a camera that senses the hand movements. A “deep learning algorithm” then process the movements and tries to pair it with American Sign Language (ASL) signs. It is able to interpret full sentences and doesn’t require the signer to stop between words. Zhang, the inventor, has applied for a patent and expects to have it on the market in about a year. The Leap Motion capture device is expected to sell for around $80.
One issue brought up by Christian Vogler from Gallaudet University, is that it may not capture all grammatical expressions used by the Deaf. Modifiers delivered via facial expressions to indicate questions, quotes, speakers…etc are likely to be lost in interpretation via AI.
Zhang admits that the technology is only capable of basic communication and has not been tested on people who are Deaf or Hard of Hearing. He does hope to keep working to remove the limitations and add facial expression recognition in order to improve DeepASL’s accuracy.
Vogler is skeptical that any AI device will ever be able to really meet the Deaf’s real-world communication needs. He suggests that efforts would be better spent in working with people in the community and interpreters in order to better provide the Deaf with better access to communication.
Vogler does suggest that the technology, which would turn hand signs into audible speech, may be a good way to interact with voice recognition devices, such as Alexa.
Grandmaster Humans
Unlike in chess, humans still rule the interpretation and translation world. AI is approaching human quality translations, but isn’t quite as accurate as we are. Unfortunately, these inaccuracies can do harm. Mistranslations are especially common in Sign Language, where additional challenges make it more difficult for AI to render a faithful interpretation. Until things improve with AI translations and interpretations, it may be in your best interest to book a professional in-the-flesh interpreter.
If you have questions or would like to book a translator or interpreter, visit our website, email us at spotoninterpreting@gmail.com, or call us at 855-562-7768 and we’d be glad to help.
Add a Comment