Abstract: In the last few years, the usage of online platforms has increased very rapidly. It lets the user to interact and upload content for others and also has become an integral part of many people's life. The user generated content posted by users contributes to the richness and variety of content on the web but isn't subject to the editorial controls associated with traditional media. Due to this some users can post content which could harm others, particularly children or vulnerable people. The amount of content generated on these platforms is increasing day by day, it's become impossible to spot and take away harmful content using traditional approaches at the speed and scale necessary. This paper examines the capabilities of machine learning technologies in meeting the challenges of moderating online content. For this we propose an Online Social Network System which does not allow its user to post hate comment or post. We have developed a machine-driven solution which automatically detects hate content in the form of text or image on social networking sites. For text classification various baseline models such as Support Vector Machine, Logistic Regression, Naive Bayes and Random Forest were used. The final model was a Support Vector Machine model that used TF-IDF for feature engineering. For image classification a combination of VGG19 pretrained on ImageNet dataset and BERT Language Model was used.
Keywords: Machine Learning, Natural Language Processing, Text Processing, Image Processing, Feature Extraction, Classification, Online Social Network, Hate, Non-Hate, Support Vector Machine, TF-IDF, VGG19, BERT
|
DOI:
10.17148/IJARCCE.2021.101009
