A Novel Approach for Predicting Nativity Language of the Authors by Analyzing Their Written Texts

Abstract

The technique of predicting the profiling characteristics like gender, age, nativity language, and location of an anonymous author’s text by examining the author’s style of writing is called Author Profiling. The differences in authors writing styles play a crucial role in Author Profiling. Several researchers extracted different types of stylistic features to discriminate the authors writing styles. Most of the researchers used a standard Bag Of Words model for document representation in Author Profiling approaches. This model has some problems like high dimensionality of features, sparsity in representation of a document and the relationship between the features were not captured in document representation. In this work, a new approach is proposed for document representation. In this approach, the weights of documents were used to generate the vectors for documents. This approach also solves the problems faced in Bag Of Words model. In this work, we concentrated on prediction of nativity language of the authors from the corpus collected from hotel reviews. Different classification algorithms were used from WEKA tool to predict the accuracy of nativity language of an author’s text. The obtained results were good when compared with the state-of-the-art approaches for nativity language prediction in Author Profiling.