TOXIC COMMENT CLASSIFICATION

TOXIC COMMENT CLASSIFICATION

Ankit Sanjyal1, Paresh Subedi2, Rohit Kumar Jaishay3, P. Manikandan4

123 B.Tech  4th Year, CSE, Jain University, Bengaluru 562112, Karnataka, India.

4Associate Professor, CSE, Jain University, Bengaluru, 562112, Karnataka, India.

E-mail: 1asanjyal56@gmail.com ; 2pareshsubedi1@gmail.com ; 3rjmentor2001@gmail.com; 4p.manikandan@jainuniver sity.ac.in ;

-----------------------------------------------------***---------------------------------------------------------------------

Abstract - In the recent times, the users on social media platforms have increased dramatically, leading to a significant increase in the amount of online content being generated. However, this surge in online activity has also led to a rise in toxicity and negativity on these platforms. Online communities are now inundated with toxic comments and hateful messages that can have serious psychological and emotional impacts on the targeted individuals. This has created a pressing need for effective methods to identify and remove toxic comments from online platforms to ensure a safe and inclusive environment for all users. To deal with this problem, machine learning models have been created to automatically classify comments as toxic or non-toxic. In this research paper, we present an approach for toxic comment classification using a Bi-directional LSTM neural network architecture. We test our approach using a publicly available dataset provided by a Kaggle competition. Our experiments show that LSTM performs excellent on text classification tasks with 98% AUC score.

Keyword’s: Machine learning, Neural Network,