Learning to Generate Popular Headlines

Headlines are not only essential for summarizing news articles but also grabbing users’ attention. Headline generation is a type of text summarization that can employ either an extractive or abstractive approach, with the latter being more prevalent through deep learning models. However, creating popular headline capture readers’ attention challenging. To address this issue, we propose hybrid approach utilizes state-of-the-art transformer models to generate several variations article. Additionally, use model predicting popularity choose most from generated ones. We create new dataset by scraping Twitter accounts media. Our evaluation shows fine-tuning task significantly improve their performance. demonstrate our proposed method headlines compared baseline methods do incorporate prediction. For such purpose, benchmark automatically assess effectiveness in generating headlines.