Fast and Accurate Resume Parsing Method Based on Multi-Task Learning

With the growing demand for online recruitment, an efficient and highly accurate resume parsing method is needed. Existing resume parsing methods are basically two-stage approaches: resume segmentation and entity recognition. Some researchers apply deep learning to these two stages to improve the accuracy but at the cost of lower efficiency. In this article, we propose a multi-task deep learning model that completes resume segmentation and entity recognition at the same time, thus reducing the resume parsing time by nearly half. Moreover, our method can better parse the dual-list format resumes, which is usually ignored by other existing methods. We also use some post-correction rules to further improve the accuracy of resume parsing. Finally, compared with the baselines, we achieve state-of-the-art on both Chinese and English resumes.