• Anoushka Samyal
• Riddhi Shinde
• Rahul Singh
• Ms. Manya Gidwani

In today’s evolving world ofsocial media,the issue of trolling and online abuse has become increasingly prevalent. Hence, it is necessary to police the toxic comments which showcase such behavior.Due to extreme online harassment and cyber bullying it has become imperative to work on models that help curb these problems. In this paper, we build an efficient model to identify and categorize toxic comments using deep learning algorithms. Deep Learning methods are starting to out-compete statistical methods on some challenging NLP problems with singular and simple models.We worked on a public dataset which is a corpus of Wikipedia comments available on Kaggle. In this work, we used word2vec + Convolutional Neural Network (CNN) approach. We tested two models, CNN with word embeddings and CNN with character embeddings. Sentimental Analysis is the field that studies and analyzes people’s responses and acceptance towards entity using text analysis computational and algorithms to help to determine people’s textual reactions if they are toxic, severe toxic, threat, identity hate, obscene or insult. We trained our dataset using keras library. We show that our CNN with word embeddings model performed with an AUC of 0.98.Here we show that CNN with word embeddings reaches the highest performance.

Convolutional Neural Networks, Deep Learning, Natural Language Processing,Sentiment Analysis, word2vec.

"TOXIC COMMENT CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORK", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 4, page no.650-656, April-2019, Available :http://www.jetir.org/papers/JETIR1904J86.pdf

"TOXIC COMMENT CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORK", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 4, page no. pp650-656, April-2019, Available at : http://www.jetir.org/papers/JETIR1904J86.pdf