An Efficient Universal Information Extraction Framework for Bid Evaluation Experts Based on Attribute-relationship

Expert information extraction (EIE) system is an automated information extraction framework designed to achieve efficient management and precise retrieval of expert resources in specific fields, whose accuracy is crucial to ensure the fairness of professional evaluation. In this research, an intelligent expert extraction system based on universal information extraction (UIE) is proposed to address the problems of high subjectivity and inefficiency of traditional manual expert extraction systems in dealing with large-scale data. Our framework uses ERNIE 3.0 as a pre-training model, and through the few-shot learning method, accurately extracts the attributes and relationships of expert information based on UIE. Thus, our system can quickly adapt and optimize the model under limited data conditions, significantly improving expert extraction’s accuracy and efficiency. Extensive quantitative and qualitative experiments have been conducted for our proposed framework, which achieves a high accuracy of 0.949 under 5-shots in the expert extraction task, significantly exceeding the performance of existing methods. The results of this research not only improve the objectivity and efficiency of EIE but also provide sustainable development of an efficient and accurate expert extraction system solution for various fields.