Tamil Author Identification Using Convolutional Neural Network

Author identification is one of the significant tasks in natural language processing(NLP) which identifies the author of a given text content. Text classification applications such as news classification and sentiment classification mostly rely on features extracted from the content or the topic of the content. Author identification, on the contrary generally depends on the features representing the writing style such as lexical, character level, syntactic and semantic features. In recent years, the performance of text classification approaches has been significantly improved by applying deep learning-based text representation. For the author identification task, a good amount of research has been done for English and other European languages whereas relatively less have been done for regional languages like Tamil. This paper describes such a study for the Tamil language. This study proposes a deep-learning model for author identification, which concatenates the conventional word-level features and automatic features learned by a convolutional neural network (CNN) with FastText pre-trained word vectors. As part of the study, a dataset for Tamil author identification was created and the proposed model was analysed. The proposed model obtained better performance compared with conventional or deep learning models.