Automated Keyword Generation for Academic Research Articles Using Large Language Models

The manual assignment of keywords in academic research articles may be subjective and inconsistent, leading to challenges in information retrieval and limiting the visibility of scholarly work. This study proposes a systematic, three-step methodology leveraging large language models (LLMs) to automate keyword generation from research articles. The proposed method utilizes three different prompts to generate keywords from article titles and abstracts, ensuring a domain-agnostic and contextually relevant approach. Furthermore, the representation vectors of LLM-generated keywords are computed and grouped based on their vector similarities. To evaluate the effectiveness of this approach, we constructed a novel dataset comprising academic articles from researchers in Türkiye, where author-assigned keywords serve as ground truth labels. Experimental results demonstrate that the Mistral model performs best within the proposed framework. Additionally, our findings highlight the significant impact of semantic grouping and prompt engineering on keyword generation success. These results point out the importance of LLM-based approaches in enhancing the accuracy and standardization of keyword assignments in academic publishing.