StructBERT中文情感分类模型介绍

from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

semantic_cls = pipeline(Tasks.text_classification, 'damo/nlp_structbert_sentiment-classification_chinese-tiny')
semantic_cls(input='启动的时候很大声音，然后就会听到1.2秒的卡察的声音，类似齿轮摩擦的声音')

import os.path as osp
from modelscope.trainers import build_trainer
from modelscope.msdatasets import MsDataset
from modelscope.utils.hub import read_config
from modelscope.metainfo import Metrics

model_id = 'damo/nlp_structbert_sentiment-classification_chinese-tiny'
dataset_id = 'jd'

WORK_DIR = 'workspace'

max_epochs = 2
def cfg_modify_fn(cfg):
    cfg.train.max_epochs = max_epochs
    cfg.train.hooks.append({
            'type': 'TextLoggerHook',
            'interval': 100
        })
    cfg.evaluation.metrics = [Metrics.seq_cls_metric]
    cfg['dataset'] = {
        'train': {
            'labels': ['负面', '正面'],
            'first_sequence': 'sentence',
            'label': 'label',
        }
    }
    return cfg

train_dataset = MsDataset.load(dataset_id, namespace='DAMO_NLP', split='train').to_hf_dataset()
eval_dataset = MsDataset.load(dataset_id, namespace='DAMO_NLP', split='validation').to_hf_dataset()

# remove useless case
train_dataset = train_dataset.filter(lambda x: x["label"] != None and x["sentence"] != None)
eval_dataset = eval_dataset.filter(lambda x: x["label"] != None and x["sentence"] != None)

# map float to index
def map_labels(examples):
    map_dict = {0: "负面", 1: "正面"}
    examples['label'] = map_dict[int(examples['label'])]
    return examples

train_dataset = train_dataset.map(map_labels)
eval_dataset = eval_dataset.map(map_labels)

kwargs = dict(
    model=model_id,
    train_dataset=train_dataset,
    eval_dataset=eval_dataset,
    work_dir=WORK_DIR,
    cfg_modify_fn=cfg_modify_fn)

trainer = build_trainer(name='nlp-base-trainer', default_args=kwargs)

print('===============================================================')
print('pre-trained model loaded, training started:')
print('===============================================================')

trainer.train()

print('===============================================================')
print('train success.')
print('===============================================================')

for i in range(max_epochs):
    eval_results = trainer.evaluate(f'{WORK_DIR}/epoch_{i+1}.pth')
    print(f'epoch {i} evaluation result:')
    print(eval_results)

print('===============================================================')
print('evaluate success')
print('===============================================================')

@article{wang2019structbert,
  title={Structbert: Incorporating language structures into pre-training for deep language understanding},
  author={Wang, Wei and Bi, Bin and Yan, Ming and Wu, Chen and Bao, Zuyi and Xia, Jiangnan and Peng, Liwei and Si, Luo},
  journal={arXiv preprint arXiv:1908.04577},
  year={2019}
}