MFE-NER: Multi-feature Fusion Embedding for Chinese Named Entity Recognition
Multi-feature embeddings for Chinese named entity recognition and character substitution.
Citation
Jiatong Li, Kui Meng. (2024). MFE-NER: Multi-feature Fusion Embedding for Chinese Named Entity Recognition. In Proceedings of the 23rd Chinese National Conference on Computational Linguistics (Volume 1: Main Conference) (pp. 1112-1122). CCL 2024.
About this work
MFE-NER combines glyph and phonetic features to recognize Chinese named entities when characters are replaced with visually or phonetically similar alternatives. Five-Stroke components represent character structure, and a phonetic encoding represents pronunciation similarity.
The features augment a pre-trained language model. Experiments evaluate character-substitution detection and overall Chinese NER performance.