19 July 2019


Photograph of microphoneCorina Newsome! This is your fault! You have to go and say:

Can we get scientific journal articles on audiobook? Please?

There is a long thread that follows about possible solutions. But two things emerge:

  1. Software to read papers aloud automatically doesn’t do a very good job.
  2. Quite a few people want these.

Following my long standing tradition of, “What the heck, I’ll have a go,” I’d like to present my first audiopaper! It’s a reading of my paper from last year on authorship disputes.

I decided to do this because I wanted to get more mileage out of a mic I’d bought for a podcast interview (forthcoming), and because I still have this discussion in the back of my head.

I often tell students, “Always plot the data”, since different patterns can give same summary stats. How could I help visually impaired students do something similar?

And the answer is that while there have been experiments in sonification of data, it seems to have stayed experimental and never moved into simple practical use. It got me thinking about how little we do for visually impaired researchers.

I picked my authorship disputes paper for a few reasons.

  1. There are no bothersome figures to worry about describing.
  2. The topic probably has wider appeal than my data driven papers.
  3. The paper is open access, so I wouldn’t run afoul of any copyright issues.
  4. The paper is reasonably short.

I wrote an little into and a little outro. I pulled out my mic, fired up Audacity, and got reading. My first problem was finding a position for the mic where I could still see the computer screen so I could read from my paper.

I broke it into sections (slightly more sections than headings the paper). I think it took between one and two hours to read the whole thing. It’s not quite a single take, but it’s close.

I’ve since figured out that I can probably do longer sessions, because I worked out how to identify sections I want to edit out because I stumbled or mispronounced words. After I screw up a sentence, I snap my fingers three times. This creates three sharp spikes in the playback visualization that is easy to see. That makes it easy to find the mistake, then edit the gaffe and the finger snaps out of the recording.

Screen shot of sound recording in Audacity comparing speech and finger snaps.

I learned that it can be surprisingly hard to say “screenplay” correctly. And I curse my past self who wrote tongue twisters like “collaborative creator credit.”

Editing the recording also took about an hour. Besides cutting out my stumbles and finger snaps, I cut out some longer pauses and occasional little background sounds. The recording was a bit quiet, so I increased the gain a few decibels.

Will I do more of these? It completely depends on the response to this experiment. I probably picked my single easiest paper to read and turn into an audio recording. It would only get harder from here. And I have other projects that I should be working on.

If people like this effort, I’ll see about doing more, maybe with better production. (I wanted to put in some music, but that was taking too long for a one off.)

External links

Resolving authorship disputes by mediation and arbitration on Soundcloud

No comments: