Smart Language Checker: A Machine Learning Solution for Offensive Language detection in Social Media

Social media serves as a prominent platform for individuals to express their thoughts, often encompassing both positive and offensive language. Less resource-rich languages benefits from automated offensive language detection systems. This article addresses this gap by proposing a model tailored for the low-resource language. Leveraging a manually labeled dataset of 60,000 comments from the dataset collected from various social media sources, the proposed model explores three feature extraction techniques: bag-of-words (BoW), Vectorization-Term frequency-inverse document frequency (TF-IDF), and Tokenization methods. Employing both traditional classifiers and a deep sequence model, the study reveals that the random forest classifier excels, achieving a testing accuracy of 94.07% when considering unigrams, bigrams, and trigrams. Furthermore, TF-IDF yields a maximum accuracy of 93.90%.The corpus generated in this endeavor is made accessible to researchers in this field, facilitating advancements in offensive language detection for English.