Clickbait detection in news headlines using RoBERTa-Large language model and deep embeddings

The integration of Large Language Models with Artificial Intelligence is transforming digital news analysis, particularly through progressions in natural language processing. Among the emerging applications, clickbait headline detection has become a significant but challenging research area. The existing research studies using Machine Learning (ML) and Deep Learning (DL) algorithms systems for news headlines analysis are limited to traditional ML and DL models. The proposed study introduces RoBERTa-Large, a transformers-based architecture, for the automated detection of clickbait news headlines. The proposed RoBERTa-Large- effectively captures complex contextual dependencies and semantic relationships within text based on its integration with self-attention mechanism. The model is evaluated against state-of-the-art ML and DL approaches to assess its classification capabilities. A diverse set of textual features including Term Frequency-Inverse Document Frequency (TF-IDF), Part of speech tagging (PoS), n-gram representations, and advanced word embeddings such as word2Vec, and FastText and Sentence Embeddings are employed to encode linguistic information from dataset. A comprehensive empirical analysis indicates that RoBERTa-Large achieves the highest classification accuracy of 97% outperforming relevant existing studies. Moreover, with Explainable AI (XAI) methods, like LIME (Local Interpretable Model-Agnostic Explanations) and SHAP (SHapley Additive exPlanations), for better understanding of the results and explainability.