Tensor-Flow Based Approach to Identify Author of the Text

Abstract

Now-a-days a lot of content is available on internet, and people upload lot of information in form of opinion, review, description, recipe etc. online. In such scenario to trace the authenticity of the data, it is necessary to develop an author identification system. It has become a difficult problem in the scope of unnamed information has increased with fast growing Internet life. It is a process to identify author of unknown text document. In existing system, so many authorship attribution methods were evaluated for natural languages, such as English, Chinese, and Arabic. Most of the experiments are done on the lexical and word based. Classification techniques such as Naïve Bayes¹,Support Vector Machine, Neural Network, Multilayer Perceptron, Decision Tree²,k-nearest-neighbor are already used, but previous work proved that Support Vector Machine³ is a good classifier for author identification. In this system we identify author using deep learning. For identification we use features like Bag-of-Words, word tokenization and stemming. We build Deep neural Network and LSTM network using the Tensor-flow library⁴. The average accuracy achieve 68 % using Tensor-flow Deep Neural Network and 38.68 % using Tensor-flow LSTM network.

Cite this Article

Kajal Patel, Brijesh S. Bhatt. TensorFlow-based Approach to Identify Author of the Text. Current Trends in Information Technology. 2018; 8(3): 23–29p.