ICT & Computing in Education

View Original

My experiments with transcribing audio

See this content in the original post

Clipart Illustration of a man speaking into a 1940s-period radio microphone. Internal Illustration from the pulp magazine Weird Tales (September 1941, vol. 36, no. 1).

Introduction

One of the few good things about fracturing my right elbow on the 21st of August 2021 was that I was finally forced into experimenting with dictating articles instead of writing them.

I was surprised, at first, to discover that dictating articles is very easy. When I thought about it though, I realised that it should not have been surprising at all. Much of the time I “write” articles in my head anyway, so dictating them was, in effect, equivalent to thinking out loud. However, just as it’s generally better to type an article and worry about formatting and spelling and so on later, it’s less of an interruption to one’s flow if, when dictating, you eschew announcing grammar and formatting commands while dictating, and deal with the layout subsequently.

My tests

I tested several applications, and I’ve included the results below. The tests were not completely commensurate with my usual blogging practice, because I dictated the text from a text I’d already created. Still, it enabled me to compare how well each application interpreted my diction, and how they responded to grammar commands.

Some findings

Things to watch out for are:

  • Incorrect spellings. Not all of the applications are good at working out the correct spelling from the context. For example you might end up with a “two” creeping in when you had intended to use “too”.

  • Some applications need you to be online for them to work properly.

  • Rogue text can slip in, such as in one of my tests when, in response to a question asked by my lady wife, I called out “Tea, please!”.

  • Some applications automatically insert commas and full stops.

The applications tested

I tried the built-in Windows 10 dictation my PC, and found it to be useless. It seemed incapable of accurately “hearing” anything I said, and I didn’t think my diction was that bad. I tried it with no mic, an expensive mic, and a really expensive mic, all to no avail. Apparently, it gets better with training, but I didn’t have the patience to persevere.

I found the same thing when I tried the built-in dictation on a Macbook Pro.

A program called Braina enables you to give voice commands to your PC computer, and to dictate text. I used the free trial version, which I found to be so cut down I was unable to try out every function.

Finally, I considered Dragon Naturally Speaking. However, the best version costs a small fortune, and there appears to be no trial version. There are less expensive versions, but they also do a lot less. Besides, apparently Dragon also needs training time in order to make it perform as well as it can.

Thus, after much deliberation and investigation, I decided to use the dictation option in the applications I already have. Here are the results. I’ve highlighted the errors in red.

Original

In the middle of the night, I woke up (if you can call being semi-conscious being awake), walked purposely towards the door to go to the bathroom — and almost knocked myself out.

The reason was that in the twin states of entire darkness and semi-somnambulance I was facing in a different direction from the one I thought I was facing. As a result, instead of walking through the door, I tried to walk through the wall.

The next few days brought nausea and headaches. After much prevarication I went to Accident and Emergency, where I waited petrified among people for whom “social distancing” means not quite touching you, and who wore their masks as a chin-warmer.

An hour and a half later I emerged into the twilight, secure in the knowledge that I had nothing more serious than mild concussion. I failed to do much writing, but I was pleased to have read a further 17% of my book.

Gmail

In the middle of the night, I woke up up.up and brackets if you can call being semi-conscious being awake) —and almost knocked myself out.

The reason was that in the twin states of entire darkness and semi sun ambulance I was facing in a different direction from the one I thought I was facing. Bold thought as a result, instead of walking through the door door all I tried to walk through the wall all.
The next few days brought nausea and headaches full stop after match prevarication I went to accident and emergency, where I waited petrified among people for whom "social distancing" means not quite touching you, and who wore them masks as a chin warmer period
An hour and a half later I emerged into the Twilight, secure in the knowledge that I had nothing more serious than mild concussion. I failed to do much writing, but I was pleased to have read a further 17% of my book.

Otter.ai

Unknown 0:02
In the middle of the night, coma, if you can call being semi conscious being awake. Walked purposely towards the door to go to the bathroom, and almost knocked myself out. The reason was that in the Twin states of entire darkness and semi Sun ambulance. I was facing in a different direction from the one I thought I was facing. As a result, instead of walking through the door. I tried to walk through the wall. The next few days brought nausea, and headaches. After much prevarication, I went to accident and emergency where I walk waited petrified among people for whom social distancing means not quite touching you. And who wore their masks as a chin warmer. An hour and a half later I emerged into the Twilight secure in the knowledge that I had nothing more serious than mild concussion. I failed to do much writing that was pleased to have read a further 17% of my book.

Colornote

In the middle of the night, I woke up up.up and brackets if you can call being semi-conscious being awake) comma walked purposely towards the door to go to the bathroom and almost knocked myself out.

The reason was that in the twin states of entire darkness and semi someone ambulance I was facing in a different direction from the one I thought I was facing as a result instead of walking through the door door I tried to walk through the wall

The next few days brought nausea and headaches. After match prevarication I went to accident and emergency, where I waited petrified among people for whom social distancing means not quite touching you and who were their masks as a chin warmer.

An hour and a half later I emerged into the Twilight, secure in the knowledge that I had nothing more serious than mild concussion. I failed to do much writing, but was pleased to have read as a further 17 percent of my book.

Google Docs

In the middle of the night, I woke up up&up and bracket tn50 can call being semi-conscious being awake) , walked purposely towards the door to go to the bathroom—and almost knocked myself out.


The reason was that in the twin states of entire darkness and semi-skimmed ambulance I was facing in a different direction from the one I thought I was facing. Bold thought as a result, instead of walking through the door come on I tried to walk through the wall.


The next few days brought nausea and headaches. After much prevarication I meant to accident and aaaaaaaaaaaaG


T FCC where I waited petrified among people for home social distancing means not quite touching you, and who all their masters as a chin warmer.


An hour and a half later I emerged into the Twilight 4koma secure in the knowledge that I had nothing more serious than my concussion. I failed to do much writing, but I was pleased to have red a further 17% of my book.


Microsoft Word

In the middle of the night, I woke up [if you can call being semi conscious being awake], walked purposely towards the door to go to the bathroom emdash and almost knocked myself out. Knew line

The reason was that the twin states of entire darkness and semi sewn ambulance I was facing in a different direction from the one I thought I was facing.

The next few days brought nausea and headaches. After much prevarication I went to accident and emergency, where I waited petrified among people for whom “social distancing” means not quite touching you, and who wore their masks as a chin - warmer.

An hour and a half later I emerged into the twilight, secure in the knowledge that I had nothing more serious than mild concussion. I failed to do much writing, but I was pleased to have read a further 17% of my book.


Conclusions

Google Docs and Gmail were very similar in their results, no doubt because they probably use the same algorithms.

Colornote wasn’t great, especially in responding to grammar instructions, but I was able to use it anywhere, with no internet connection.

Otter.ai was pretty good, especially as it inserted punctuation marks automatically. The only downside was that it exported the text into one big paragraph, but that’s easily sorted out.

The one I found by far the best was Microsoft Word, which interpreted grammar commands very well, and understood what I was saying, on the whole. The only downside is that the dictation only works, I think, when you’re online.

A footnote

After fracturing my elbow I had intended to write this blog post much sooner. However, family issues intervened, and now is the first chance I’ve had to put it all together. In the meantime, Stephen Downes reported on transcribing and includes tools I haven’t mentioned here, and provides a link to the original audio so you can see for yourself how accurate the transcription was.