Authorship detection model Using descriptive analysis

Abstract

Writing skills are very important for succeeding as a writer. Every writer has specific style and way of communicating his thoughts. This specific way can be reusing specific set of vocabulary again and again, focusing on a specific point and pro or against an idea. However, these specific characteristics of authors may change from one domain to another i.e. sports, politics and entertainment etc. Considering this fact, the current research was conducted to automatically identify the author of a particular piece of text. An approach is developed which automatically suggests names of authors for a particular piece of text considering the features of authors. For this study, a set of 16 article writers have been taken, which are very well-known in their field of writings. Furthermore, twenty articles written by each author were included in the training data set. After selection of the specified set, we collected a number of articles for testing the proposed algorithm. By providing those specified articles, we train our algorithm and then hy providing anonymous articles to our algorithm for testing and evaluation purposes. The algorithm suggests top three authors (out of 16) for a given piece of text. Which depicts the similarity between each article of the specified author. Our algorithm is based on few measures like cosine-similarity, ranking, precision, accuracy and F-measure. The evaluation of the algorithm shows that the accuracy of the proposed algorithm is 83%.