Transformer-based pre-trained language models (e.g. ChatGPT) have made tremendous progress in the fields of natural language understanding and generation. In the domain of text summarization, both extractive summarization (BERT, RoBERTa) and abstractive summarization (BART, T5) models have achieved remarkable performance. This SURF project focuses on developing a summarization model that can generate accurate and concise summaries of scholarly articles. However, current summarization models may generate factually inconsistent text with respective of their inputs, which can lead to misinformation. To address this issue, we plan to extract important facts from the original text in the form of triples and use external knowledge bases such as knowledge graphs to improve the authenticity of the generated content. We aim to enhance the efficiency of the scholarly research process by providing researchers with a tool to quickly and accurately identify relevant articles and key points of interest.