A Chinese MRC model built on Chinese PERT-large
Please use BertForQuestionAnswering to load this model!
This is a Chinese machine reading comprehension (MRC) model built on PERT-large and fine-tuned on a mixture of Chinese MRC datasets.
PERT is a pre-trained model based on permuted language model (PerLM) to learn text semantic information in a self-supervised manner without introducing the mask tokens [MASK]. It yields competitive results on in tasks such as reading comprehension and sequence labeling.
Results on Chinese MRC datasets (EM/F1):
(We report the checkpoint that has the best AVG score)
| CMRC 2018 Dev | DRCD Dev | SQuAD-Zen Dev (Answerable) | AVG | |
|---|---|---|---|---|
| PERT-large | 73.5/90.8 | 91.2/95.7 | 63.0/79.3 | 75.9/88.6 |
Please visit our GitHub repo for more information: https://github.com/ymcui/PERT
You may also be interested in,
Chinese Minority Languages CINO: https://github.com/ymcui/Chinese-Minority-PLM
Chinese MacBERT: https://github.com/ymcui/MacBERT
Chinese BERT series: https://github.com/ymcui/Chinese-BERT-wwm
Chinese ELECTRA: https://github.com/ymcui/Chinese-ELECTRA
Chinese XLNet: https://github.com/ymcui/Chinese-XLNet
Knowledge Distillation Toolkit - TextBrewer: https://github.com/airaria/TextBrewer
More resources by HFL: https://github.com/ymcui/HFL-Anthology