Generating Headlines from Article Summaries Using Transformer Models

Generating headlines for articles requires a deep understanding of the article's content and context. Automating the process of generating titles for articles from their summaries can save significant time and effort for journalists and other content creators. The primary objective is to automate the process of generating titles for research papers or articles by using Natural Language Processing (NLP) techniques and Machine Learning models. The study uses an open-source dataset from Kaggle with nearly 50000 records., each featuring titles., summaries., and terms. Various NLP techniques., such as tokenization and TF -IDF vectorization were employed to process and analyze the textual data. The proposed approach involves transfer learning based on pre-trained transformer architectures like BART and SimpleT5., as well as machine learning models., including SVM., K-Means., and Decision Trees. The result is successfully generated automatically with concise and informative titles that accurately represent the content of the article. Out of all the models proposed., BART has shown promising results with a minimal evaluation loss of 0.05.