Detection of offensive text in memes using deep learning techniques.

Memes have grown in significance as social media has grown in popularity. Memes are challenging to categorizeusing conventional techniques since they are typically constructed using a combination of images and text, and their contentis frequently hilarious or sarcastic. We suggest a deep learning-based strategy in this paper for classifying memes into a variety of categories, such as sexism, politics, and criticism. This highlights the necessity for a system that can evaluate memes automatically before they raise controversy or spread humor. This study offers a three-step method for locating offensive memes. Before judging the text as offensive or not, it will first extract the text from the supplied image. If the language is found to be unacceptable, the third phase will further categorize the information into three categories: mildly offensive, very offensive, and hateful offensive. The five thousand memes that made up the dataset for this study were divided into four categories: hateful offensive, highly offensive, slightly offensive, and not offensive at all. Concern over the dissemination of offensive memes on social media has grown in recent years. Memes that are offensive can hurt and distress people individually as well as in communities since they frequently contain graphic or disparaging content. Due ofthe complexity of language and the range of cultural allusions used in memes, it can be difficult to identify and remove offensive memes. In this study, we provide a CNNs, Roberta, and BERT-based deep learning approach for categorizing offensive memes. We enhanced pre-trained algorithms using a dataset of offensive and non-offensive memes. We evaluatedthe performance of each model using various evaluation metrics, such as accuracy, precision, recall, and F1-score. [ABSTRACT FROM AUTHOR]