Microsoft Tay

How I built it

Implemented simple MCMC algorithm on English text gathered from reddit (/r/ Philosophy) and Donald Trump's website to predictively generate text given a keyword of phrase.

Challenges I ran into

My original aim was to use a framework such as TensorFlow to train a deep learning model (as microsoft tay, xiaoice, etc.) My quota on the CSUG servers was too little, and the resulting model generated poor text (repeat single character words). With a larger dataset, quota, and more time, I could have finished this approach or improved the output of the current.

What's next for mcmc_philosoph_trump

Collect more data

Built With

Share this project: