Uses in #Academia :
I am beyond impressed with how #ChatGPT assisted me this afternoon to get #Whisper running on my Mac.
I have some 20+ hours of potato quality interview recordings for a book project to transcribe. First trials with open Whisper services looked promising, but I didn't want to sign up with a commercial service. So I decided to try and run it on my own machine.
There is a nice jupyter tutorial on Github, posted by lablab.ai – but it is for windows, I haven't updated python since 2.7, nor written anything significant for more than five years. And I've actually never used jupyter. Sounds bad, right?
Well ChatGPT walked me through it all step by step, reminded me of trivial concepts that I had long forgotten, helped me downgrade to python 3.10 when 3.11 turned out to be unsupported, made sure my pips and paths got untangled, wrote a timing command for me... It took two hours ... but I know I would not have been able to solve this at all otherwise, and gone with the commercial option instead.
That we have a computer program that can help with this level of tutoring for beginner students is an absolute game changer.
Now my CPU is happily churning away, spitting out an hour's worth of transcripts (In amazing quality I must say) each ten minutes, and my sanity is safe. (For now anyway ...)
🙂
#academia #whisper #asr #AutomaticSpeechRecognition #chatgpt #highereducation
#introduction My name is Matt, and I am a phonetician and postdoc in the Department of Linguistics at the #UniversityOfWashington. I am also the currently the acting director of the UW Phonetics Lab.
In #phonetics, my research focuses on bringing more acoustic information into #SpokenWordRecognition and #LexicalModeling. I am also interested in #SpeechTechnology such as #ForcedAlignment and #AutomaticSpeechRecognition methods.
#introduction #universityofwashington #phonetics #SpokenWordRecognition #LexicalModeling #SpeechTechnology #ForcedAlignment #AutomaticSpeechRecognition