A previous study on website of one of our team members (https://smunoz.dev/) inspired him to make:
- A rapping bot to answer the rhymes of the user with structure, semantics, and rhyme.
- Complete rewrite of Word2Vec study with an english dataset instead.
What it does
- Bot answers to your phrases over:
- Web app using voice recognition, answers with text, user can leave ratings. (vue_frontend)
- User writes on console, gets an answer over the phone with TTS. (markov/run_twilio)
- User writes on console and gets an answer back over text in the console. (markov/run.py)
- Twitter bot (not finished)
What it doesn't do
We trained some markov models, and Word2vec, but we weren't able to combine them because markov has no memory of previous states of large scope. At least it finds rhymes. We downloaded a pre trained GPT-2 model to plug it into a twitter bot, but we wanted to train it with a bigger dataset, but we had no time.