Recording verbal speech while alseep, transcribing words to text and running sentiment analysis on the whole thing

I recently started a new Quantified Self related project that I think has some real potential for (1) revealing my underlying health state and (2) being replicated by anyone else interested in trying it for themselves; and for cheap. To keep things simple, I’m recording my sleep-talk and will eventually run sentiment analysis on enough data as to highlight my overall wellness.

To begin this quasi-experiment, I’ve purchased a microphone and recorder that I will use to gather audio files from 30+ consecutive nights of sleep. I will then edit that audio down to just the parts where I’m talking aloud while sleeping. Next, using a speech-to-text service, I plan to transcribe all of those mumbled words into a body of data. Finally, I will take all of the text that I’ve generated, and push it all through a custom built sentiment analysis tool to see what I can find.

From this moment in time, I’m assuming I’ll amass enough text data after a month or so. Maybe two months. And with the power of AI, I’ll be able to transcribe all of that audio into text effortlessly. All of this can easily be accomplished by anyone for less than $60 USD. I’m excited to update this community with my progress as I move forward.

I could also use a text analysis tool (such as this one) to help digest what I end up saying while asleep.

2 Likes