Enhancing Academic Title Drafting Through Abstractive Summarization

The title of an academic paper encapsulates its core knowledge at a high level, while the abstract provides a concise summary of the paper's content. Therefore, an automatic title generation model utilizing abstractive summarization techniques can significantly aid researchers in drafting titles. This paper presents a model for generating titles for Chinese academic papers, leveraging the Transformer architecture and the pre-trained BERT model. Experimental results demonstrate that the Transformer model, augmented with pre-training, outperforms other models, achieving a Rouge-L score of 0.441. Human evaluations further indicate that the generated titles effectively support the decision-making process in academic title drafting.