LLM Collaboration PLM Improves Critical Information Extraction Tasks in Medical Articles

With the development of modern medical informatics and databases, medical professionals are increasingly inclined to use evidence-based medicine to guide their learning and work. Evidence-based medicine requires a large amount of data and literature information, where most search processes are keyword retrieval. Therefore, anticipating these key information through the model can play an important role in optimizing the query. In the past, the PLM (Pre-trained Language Model) model was mainly used for information extraction, but due to the complexity of the sequence semantic structure and task diversity, it is difficult for traditional PLM to achieve the desired effect. With the advancement of LLM (Large Language Model) technology, these issues can now be well managed. In this paper, we discuss the information extraction evaluation task CHIP-PICOS, and finally decompose it into classification and information extraction sub-problems, applying PLM and LLM respectively, and analyzing the advantages and disadvantages and differences between PLM and LLM. The results show that our framework has achieved significant performance.