[Emc2fma]

I agree completely. Cultivating the discipline to maintain a daily journal is the hardest part. Could I suggest that you check out https://www.60secondseveryday.com?

It's the fastest way of daily journaling - you make a 60 second phone call every night to answer the question "What do you want to remember about today?". From there, it gets transcribed, archived, and it's all searchable as well.

Disclaimer: I came up with the service after having a need for it myself

[abrichr]

Cool! Which STT library/service are you using? How are you finding the error rate?

[Emc2fma]

Glad you like it! Have you had a chance to try it out yet?

Currently I'm using Google's Speech API. Based on my tests, they have the best transcription results by far.

I experimented with a few others (IBM, for one) and found that while the coding was easier (because they allow a wider variety of audio file formats), the results weren't as accurate. Google, on the other hand, is fairly picky about what types of files can be processed - this is also probably what makes them more accurate. So in the background, there's a fair bit of audio transcoding that goes on before transcription.

What do you mean by error rate?

[abrichr]

I haven't tried it out yet.

By error rate I mean Word Error Rate (WER): https://en.wikipedia.org/wiki/Word_error_rate

[Emc2fma]

I can't give you an exact number because the audio/transcripts for users are completely hidden from human eyes/ears.

Based on my own personal experience however, I would say I've gotten around 85% accuracy.

[TheOneTrueKyle]

Oh neat. I might have to give this a try. Thanks!

[Emc2fma]

You're welcome!

Have you had a chance to have your first call yet?