Glyce: Glyph-vectors for Chinese Character Representations

NeurIPS 2019 Yuxian MengWei WuFei WangXiaoya LiPing NieFan YinMuyu LiQinghong HanXiaofei SunJiwei Li

It is intuitive that NLP tasks for logographic languages like Chinese should benefit from the use of the glyph information in those languages. However, due to the lack of rich pictographic evidence in glyphs and the weak generalization ability of standard computer vision models on character data, an effective way to utilize the glyph information remains to be found... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Chinese Word Segmentation AS Glyce + BERT F1 96.7 # 1
Precision 96.6 # 1
Recall 96.8 # 1
Chinese Sentence Pair Classification BQ Glyce + BERT Accuracy 85.8 # 1
F1 85.5 # 2
Precision 84.2 # 1
Recall 86.9 # 1
Chinese Dependency Parsing Chinese Pennbank Biaffine + Glyce LAS 89 # 1
UAS 90.2 # 1
Chinese Sentence Pair Classification ChnSentiCorp Glyce + BERT Accuracy 95.9 # 1
Chinese Word Segmentation CITYU Glyce + BERT F1 97.9 # 1
Precision 97.9 # 1
Recall 98 # 1
Chinese Semantic Role Labeling CoNLL-2009 k-order pruning + Glyce F1 83.7 # 1
Precision 85.4 # 1
Recall 82.1 # 1
Chinese Part-of-Speech Tagging CTB5 Glyce + BERT F1 96.61 # 2
Precision 96.5 # 1
Recall 96.74 # 1
Chinese Part-of-Speech Tagging CTB6 Glyce + BERT F1 95.41 # 1
Precision 95.56 # 1
Recall 95.26 # 1
Chinese Part-of-Speech Tagging CTB9 Glyce + BERT F1 93.15 # 1
Precision 93.49 # 1
Recall 92.84 # 1
Chinese Sentence Pair Classification Fudan corpus Glyce + BERT Accuracy 99.8 # 1
Chinese Sentence Pair Classification iFeng Glyce + BERT Accuracy 87.5 # 1
Chinese Sentence Pair Classification LCQMC Glyce + BERT Accuracy 88.7 # 1
F1 88.8 # 1
Precision 86.8 # 1
Recall 91.2 # 1
Chinese Word Segmentation MSR Glyce + BERT F1 98.3 # 2
Precision 98.2 # 1
Recall 98.3 # 1
Chinese Named Entity Recognition MSRA Glyce + BERT F1 95.54 # 4
Precision 95.57 # 1
Recall 95.51 # 1
Chinese Sentence Pair Classification NLPCC-DBQA Glyce + BERT F1 83.4 # 1
Precision 81.1 # 1
Recall 85.8 # 1
Chinese Named Entity Recognition OntoNotes Glyce + BERT Precision 81.87 # 1
Recall 81.4 # 1
F1 80.62 # 1
Chinese Word Segmentation PKU Glyce + BERT F1 96.7 # 1
Precision 97.1 # 1
Recall 96.4 # 1
Chinese Named Entity Recognition Resume NER Glyce + BERT F1 96.54 # 1
Precision 96.62 # 1
Recall 96.48 # 1
Chinese Part-of-Speech Tagging UD1 Glyce + BERT F1 96.14 # 1
Precision 96.19 # 1
Recall 96.1 # 1
Chinese Named Entity Recognition Weibo NER Glyce + BERT F1 67.6 # 2
Precision 67.68 # 1
Recall 67.71 # 1
Chinese Sentence Pair Classification XNLI Glyce + BERT Accuracy 79.2 # 1