A Stylometric Approach for Author Attribution System Using Neural Network and Machine Learning Classifiers
Anika Samiha Hossain, Nazia Akter, Md. Saiful Islam · 2020
The paper discusses few methods of author attribution system in Bengali literature through stylometric approach. Our goal is to examine whether it is possible to identify the actual writers of some unknown Bangla documents by using a machine learning algorithm and artificial neural network. Two voting systems are also generated by the algorithms and how it varies from the result of classification models is discussed here too. We have made a corpus collecting articles of eight political writers. Firstly, we have selected some effective style markers based on statistical analysis of extracted features. Then multilayer feedforward neural network and SVM classification model are used to build an attribution system. Later, two voting systems are created by MLP classifier and SVM classification. By doing experiment and comparison, voting system gives much better result and works in more effective way for our research than the classification models.