We often talk about translation technology on this blog and how it compares to human translators. We’re not the only ones either, it’s a huge topic in our industry and something many tech firms invest big money into.
There’s a similar debate that doesn’t get the same kind of attention though. People have been using transcription software to turn audio into text since the early 1980s, and the technology has come a long way since then. Countless industries now rely on machine and human transcription to document essential information. So how good is the transcription technology on offer today?
What exactly is transcription technology?
Transcription is the practice of turning human speech into text, although that definition makes it sound like a single process, when it’s actually two rolled into one. First, you have to recognise the speech you want to document and then you can convert it into text. So transcription technology actually does two things: voice recognition and text conversion.
This is important because machines are at a disadvantage to humans when it comes to speech recognition. We’re pretty good at interpreting speech until new languages come into the mix. However, we can still come unstuck with low-quality recordings or fast speech, for example. These are challenges we have to overcome in our own transcription services from time to time.
The point is, once you interpret the speech, converting it to text is relatively simple, and the same goes for transcription technology. In fact, it’s easier for machines because they can produce large volumes of text in no time at all. This is the main strength of transcription technology: speed.
What are the weaknesses of transcription software?
As you can guess by now, accuracy is the biggest weakness of transcription software. It has this in common with translation technology but not quite to the same extent. Transcription software still has to interpret the meaning of source material but it doesn’t have to convert it into another language.
One answer to today’s question is that transcription technology is generally more accurate than machine translation. However, factors like accents, speech clarity, pronunciation and various other voice-recognition challenges pose similar problems to transcription software. There’s also the question of context to consider too. Transcription software can’t interpret sarcasm, emotion or the deeper meaning of speech, so it comes with its own limitations.
When should I use transcription technology?
As far as transcription technology has come over the years, it’s not yet close to handling projects on its own. You’ll have seen this yourself anytime you play around with Siri or Google Now’s voice recognition, which can be hit and miss to say the least. However, a good piece of transcription software can be incredibly useful for larger projects, as long as it can interpret the source material with a good level of accuracy.
In these cases, a human transcriber can use the software to transcribe the bulk of a project and take more of an editing role. This isn’t suitable for every type of project but transcription technology will be able to take on more responsibility as it continues to improve. So, much like machine translation, transcription technology can be a powerful tool if it’s used in the right settings. Just don’t trust all your transcription needs to software without a trusty human on call for quality control!