Authorship Attribution using Tf-Idf weight with Machine Learning Approaches

Authorship attribution means automatically determining the authorship of an unrecognize document based on a given sample of text. Authorship attribution has a long history and a wide range of applications. As digitalization is being increased Authorship attribution is playing a vital role in forensics, plagiarism detection, authorship dispute, and research fields. It is very helpful when two persons claim ownership of the same document written by him/her. Authorship attribution is a classification problem, although it uses the techniques of text classification for the text pre-processing despite that it is far different from the goal of text classification because it purely depends on the non-deterministic writing style of the author. The success of the author attribution task highly depends on the writing style characteristics of the authors. Several researchers proposed many different kinds of features such as linguistic, syntactic, semantic, and content-based features to identify the writing style of the authors and many of them have been used by different classifiers to classify the documents. In this paper, we applied tf-idf frequency with FW and sylometric features with support vector machines and parametric and nonparametric methods with supervised and unsupervised classification techniques in authorship attribution. We conducted various experiments with the English corpus gathered from PANCLEF'12. We performed experiments on different feature sets extracted from Corpus with different classifiers and combined these results to improve our success rates. We identified which classifiers give good results on which feature sets. According to experiments, the success rates dramatically change with different combinations of feature sets, however, the test among them are support vector classifier with a bag of words, and Gaussian with function words.