Inspiration
What it does
Automated speech-to-text transcription and easy real-time editing.
How we built it
Separately, we implemented the Vosk speech-to-text, BERT context algorithm, and the web page. Then, we merged the side projects into the final product.
Challenges we ran into
Vosk prints out words in a sentence-by-sentence fashion rather than word-by-word, detrimental to our plan to combine it with BERT. Moreover, BERT and Vosk only output the best option; however, the goal was to have a list of alternatives. So, a big roadblock was changing the code in both algorithms to generate the top few alternatives, not just the single best.
Accomplishments that we're proud of
Putting it together! None of us were very familiar with Vosk, BERT, or web design so the amount we have learned during the hackathon is an accomplishment itself.
Log in or sign up for Devpost to join the conversation.