Manchester United manager Alex Ferguson has had his words mangled by YouTube’s auto-captioning system in an example that proves human transcription always beats the machine equivalent. YouTube’s captions are used by deaf and hard of hearing viewers so they can follow the videos on the Google-owned site. Speaking ahead of the first part of Man U’s Champions League battle with Real Madrid earlier in March, Ferguson attended a press conference that was recorded and made available on the site. However, his talk of strategy and predictions were somewhat lost on any viewers using the caption tool. Lost in transcription Fergie told the gathered reporters: “It’s two clubs, two teams who embrace the best parts of football. Great reputations and, of course, Real have won the trophy nine times.” Yet the caption reads: “Transposon jumbotrons also MagicJack crabs three times and Chick Webb’s commerce about spousal travel prohibition on the past.” Then, Fergie’s quote “I think it’s a bit beyond me for the rest of my career but you never know. So, we’re looking forward to it”, was transcribed as: “We are on Monday cover many times having some of the money for Melissa McElroy fiddle slowed in front of it.” This is not the first time YouTube’s transcription tool has struggled with the Scottish accent. According to the Daily Record, the video hosting site has also mistranscribed the words of first minister Alex Salmond, along with comedian Kevin Bridges. While the gobbledygook content the transcription tool comes out with may be amusing to many, it is not regarded this way by those who rely on it. Paul Redfern, business development manager for the British Deaf Association, told the newspaper: “The problems with YouTube are symptomatic of a larger problem.” He added that “lots of services on the internet are not accessible to deaf people”. A spokeswoman for YouTube responded by saying that the technology transcribed videos in real time but worked best on “clearly spoken American-English”. Yet very few people speak with a standardised American-English accent. YouTube has said it is working to improve its captioning technology but cases like this serve to highlight the benefits of employing a human transcriptionist. While regional accents, as well as problems with audio quality and background noise, can prove a challenge to a human transcriptionist too, they have the intelligence to deal with it. So, when transcribing Fergie’s interview, for example, they know he is Manchester United’s manager and that the press conference is taking place ahead of a Champions League game so they will be able to use common sense to work out any words they can’t hear clearly or when his accent trips them up. A human will also be much better at understanding a Scottish accent – or any other strong dialect – than a machine. Tackling translation Given the difficulties YouTube has had with its caption tool, it’s hard to feel much confidence about the effectiveness of its translation tool. In September, it was announced that YouTube had introduced a new tool to its video editor, which allows users to translate the captions added to their videos into as many as 300 languages. Users can either crowdsource and allow other members to translate the content into their particular mother tongue or use Google Translator Toolkit editor’s new automated caption translator. However, looking at YouTube’s recent track record when it comes to automated transcription, choosing the automated translation could be a risk. For this reason, it’s vital any user carefully considers what they are trying to achieve with their video. A fun home video will not suffer if the automated caption option is chosen. However, if it is an important business video designed to inform clients or attract new customers in an overseas market, it’s important a professional human translator is chosen to supply the captions. Similarly, if the film requires captions in English, it is worth getting a professional transcript rather than risking the machine version. As Tech Spot’s Rick Burgess said last September: “Given how bad some of YouTube’s same-language, machine-generated captions have proven to be, one can only imagine the kinds of funny results we’ll see by making poor translations of bad transcriptions.” Fergie may certainly wish a human had produced his captions.
