Authorship attribution of short texts using multi-layer perceptron

Authorship attribution using stylometry techniques to analyse texts has grown out from earlier times for verifying the authenticity of evidence, authorial identity among other things. With the advent of the digital era, traditional pen paper writing is replaced by electronic documents making earlier techniques of handwriting analysis impossible because their electronic nature eliminates the informative differences in authorial style. Previously, authorship attributions focused mainly on unmasking the author of long pieces of digital texts but in this study, we are going to do the same for short texts that are shared on social platforms and boards. We have used a multi-layer perceptron to correctly attribute short texts to their authors using a Twitter dataset of four authors and 400 tweets for each author with 96.44% accuracy.