With the proliferation of digital content, efficient text classification algorithms are becoming more and more important. This paper presents a novel framework for convolutional neural network (CNN)-based comment classification in response to the growing social impact of online abuse. To build and assess our model, we experimented with data from the Kaggle dataset. The suggested architecture uses deep neural network techniques to automatically extract important elements from text and consists of embedding, convolutional, pooling, and dense layers. Metrics such the F1 Weighted Score, ROC AUC OVR Score, and hamming score were used to evaluate performance. The training and testing hamming scores were 0.98 and 0.97, respectively. The outcomes indicate how well the neural network model learns from data and how well it functions with invisible inputs, highlighting its potential to improve online content.