Skip to content

Latest commit

 

History

History
45 lines (38 loc) · 1.13 KB

README.md

File metadata and controls

45 lines (38 loc) · 1.13 KB
Raw

ChnSetiCorp task

ChnSetiCorp is a dataset for sentiment analysis.

Dataset

The corpus can be find HERE
Download the corpus and save data at [ChnSetiCorp_DATA_PATH]

Train and Evaluate

Download ChineseBERT model and save at [CHINESEBERT_PATH].
Run the following scripts to train and evaluate.

python ChnSetiCorp_trainer.py \
  --bert_path [CHINESEBERT_PATH] \
  --data_dir [ChnSetiCorp_DATA_PATH] \
  --save_path [OUTPUT_PATH] \
  --max_epoch=10 \
  --lr=2e-5 \
  --warmup_proporation 0.1 \
  --batch_size=16 \
  --weight_decay=0.0001 \
  --gpus=0,

Result

The evaluation metric is Accuracy.
Result of our model and previous SOTAs are:

base model:

Model Dev Test
ERNIE 95.4 95.5
BERT 95.1 95.4
BERT-wwm 95.4 95.3
RoBERTa 95.0 95.6
MacBERT 95.2 95.6
ChineseBERT 95.6 95.7

large model:

Model Dev Test
RoBERTa 95.8 95.8
MacBERT 95.7 95.9
ChineseBERT 95.8 95.9