Chinese resume multi-entity recognition method based on BERT-BiLSTM-CRF combined model
By using the BERT-BiLSTM-CRF model, combined with a pre-trained language model and a bidirectional long short-term memory network, the problems of entity type ambiguity and inaccurate boundary localization in Chinese resumes were solved, achieving high-precision multi-entity recognition and improving the recognition effect.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JINBAOXIN SOCIAL SECURITY CARD TECH CO LTD
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-03
AI Technical Summary
Existing technologies for processing Chinese resumes suffer from problems such as ambiguity in entity types, insufficient capture of long-distance dependencies, and inaccurate entity boundary localization, resulting in low recognition accuracy and recall.
The BERT-BiLSTM-CRF joint model is adopted. A deep semantic vector sequence is generated by a pre-trained language model and offset mapping information is recorded. A bidirectional long short-term memory network is combined to capture long-distance contextual dependencies. A conditional random field layer is used for decoding, and entity boundaries are aligned based on the offset mapping information to achieve high-precision entity recognition.
It improves the accuracy and recall of entity recognition in Chinese resumes, accurately distinguishes entity types, solves the problem of entity boundary positioning offset, and meets the strict requirements of downstream applications for location information.
Smart Images

Figure CN122334248A_ABST