Speech to text services

August 23, 2011 at 2:46 pm | Posted in systems | Leave a comment
Tags:

It’s over a year since I blogged about text-to-speech, when my conclusion was that automated conversion was insufficiently accurate to be useful. A few weeks ago I was contacted by a company called dictate2us which offers a web-based speech to text service, this time based on human transcription. You create an account, add some funds (I used PayPal) and upload your audio file to their website. Turn-around time depends on the length and complexity of the recording, but can be less than an hour for 5 min or less from a single-speaker. The cost is around £1 per minute of audio, so it pays to speak relatively quickly! There is an  iPhone app that effectively works like a combined dictaphone and typist – just the thing for a businessman on the move. You can even upload Word templates that they will slot your text into, so that your letters are ready to send as soon as you’ve given them a quick once-over.

I uploaded the MP3 audio from one of my shorter Panopto recordings as a test – the idea being to see how simple it would be to create fully accessible recordings using Synote. The dictate2us service was smooth and easy, and the text returned within the promised deadline. You do need to review it – there were a few small errors, some of them due to my poor diction and others introduced by the transcriber. For example they missed out the word ‘JISC’ – I said it clearly enough, but it means nothing to most people and so it was omitted. There might be similar problems with transcriptions of many academic subjects that involve specialised vocabulary. Overall, though, accuracy was high and the spelling and punctuation was good.

Of course the chief problem from an academic perspective is funding. It might ‘only’ cost £50 to transcribe a lecture, but who pays, and is it scalable to a module or even a programme?

The next steps are to test a couple more five-minute chunks of  ‘real’ lectures with specialised content – and also to see how much effort it takes to synchronise these transcripts with Synote.

Advertisements

Leave a Comment »

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Blog at WordPress.com.
Entries and comments feeds.

%d bloggers like this: