Search for the real McCoy: Authorship Attribution

A person's writing style is an example of a behavioral biometric identity. The words used by certain individuals, distinctive structuring of the sentences, personification of vocabulary, writing discipline and articulation of theory can often be used to link a piece of written work. Authorship attribution is one of the oldest problems in linguistics which also became one of the most important ones to be solved with the rise of modern statistics, machine learning and natural language processing. Authorship attribution is defined as an attempt to identify if the testing corpus has been written by the aforementioned author or not using their stylometric fingerprint. Here the proposed research discusses in detail about solving this challenging problem of identifying patterns in text using multiclass classification techniques coupled with Radial Basis Function. The proposed model yields state-of-the-art performance on several data sets, containing either formal texts written by a closed set of authors or informal texts generated by thousands of online users. Further discuss on the applicability of such novel algorithm in cases like email authorship verification and others are also carried out. Finally the findings have been formulated as a set of recommendation for best practices.