Topological Data Analysis for Distinguishing Human Written and AI Generated Abstracts

Distinguishing human written text from machine generated text is a widely studied topic. In our approach we intend to apply this to purely academic text, and compute Vietoris-Rips up to H2 to produce persistence diagrams which we can statistically analyze. We contribute a classification process with promising success rates. Later, we analyze the transformer structure using Alpha complex filtration.