Priyanshu Nayan

Text classification with Fasttext

I wanted to make a project where I could predict whether a text is computer-generated or human-generated. I immediately knew this was a binary classification problem. I only knew logistic regression, but now I also know that better alternatives exist. So I started looking for good algorithms for binary text classification and stumbled upon fasttext. Fasttext is a library by Meta. Its precision and recall were 0.90, which is not bad. I will explore if there are other better possibilities. The dataset I used was this. This is how you can get fasttext working quickly:

I post more frequently on Twitter. Follow me there to see what I am upto these days!