Human Resource Recommendation Using Chinese BERT, BiLSTM, and Short Text Similarity Algorithm

Abstract

Nowadays, many job seekers often cannot clearly analyze their job advantages and the real needs of the position, which can easily lead to large job seeking deviations, low admission rates, and waste of talent in the human resources market. The study first uses character embedding technology to pretrain Chinese text. Secondly, to enhance the model’s ability to understand contextual information in text, a bidirectional long short-term memory network and attention mechanism are introduced. Finally, the short text similarity algorithm is used to calculate the matching degree between job seekers and job descriptions. The experimental outcomes denoted that the data classification accuracy of the new model reached a high of 97.3%, which was about 5% higher than that of the Chinese BERT module alone. The highest recommendation success rate was 97.6%, the highest job recommendation acceptance rate was 97%, the highest job matching degree was 97.83%, the lowest average processing time was 3.28 seconds, and the highest user satisfaction was 98.88%. From this, the model proposed by the research has excellent performance in occupational data classification and human resource recommendation among many existing models, and can provide an effective technical support for subsequent human resource recommendations and market operations.