---
title: nlp_bert_sentence-similarity_english-base
canonical_url: "https://www.modelscope.cn/models/iic/nlp_bert_sentence-similarity_english-base"
md_url: "https://www.modelscope.cn/models/iic/nlp_bert_sentence-similarity_english-base.md"
repository: iic/nlp_bert_sentence-similarity_english-base
chinese_name: "BERT文本相似度-英文-base-学术数据集paws"
last_updated: 2025-06-11
license: "Apache License 2.0"
pipeline_tag: sentence-similarity
tasks:
  - sentence-similarity
model_type:
  - bert
library_name:
  - pytorch
frameworks:
  - pytorch
language:
  - en
domain:
  - nlp
downloads: 1570
stars: 5
tags:
  - transformer
  - Alibaba
  - sentence-similarity
---

# nlp_bert_sentence-similarity_english-base

> nlp_bert_sentence-similarity_english-base - iic 在 ModelScope 开源的模型。该模型是在bert-base-uncased预训练模型的基础上，用paws数据集训练出来的文本相似度匹配模型。

iic/nlp_bert_sentence-similarity_english-base 是 ModelScope 魔搭社区上的sentence-similarity模型，采用 Apache License 2.0 许可。

- **Repository**: iic/nlp_bert_sentence-similarity_english-base
- **License**: Apache License 2.0
- **Tasks**: sentence-similarity
- **Tags**: transformer, Alibaba, sentence-similarity
- **Downloads**: 1570
- **Stars**: 5
- **Last updated**: 2025-06-11

Source: https://www.modelscope.cn/models/iic/nlp_bert_sentence-similarity_english-base

---

# BERT文本相似度-英文-base-学术数据集paws

该模型基于bert-base-uncased，在paws(Paraphrase Adversaries from Word Scrambling)数据集(约9万条)上微调得到。

## 模型描述
![模型结构](model.jpeg)

模型采用英文句子对方式对句子相似度进行学习

## 期望模型使用方式以及适用范围
你可以使用BERT英文文本相似度模型，对通用领域的英文文本相似度任务进行推理。
输入形如（文本A，文本B）的文本对数据，模型会给出该文本对的是否相似的标签（0, 1）以及相应的概率。

### 如何使用
在安装完成ModelScope-lib之后即可使用，请参考[modelscope环境安装](https://modelscope.cn/docs/%E7%8E%AF%E5%A2%83%E5%AE%89%E8%A3%85) 。

#### 推理代码范例
```python
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

semantic_cls = pipeline(Tasks.sentence_similarity, 'damo/nlp_bert_sentence-similarity_english-base', model_revision='master')
semantic_cls(input=('That is a happy person', 'That person is happy'))

```

[//]: # (#### Finetune/训练代码范例)

[//]: # (```python)

[//]: # (import os.path as osp)

[//]: # (from modelscope.trainers import build_trainer)

[//]: # (from modelscope.msdatasets import MsDataset)

[//]: # (from modelscope.utils.hub import read_config)

[//]: # ()
[//]: # ()
[//]: # (model_id = 'damo/nlp_bert_sentence-similarity_english-base')

[//]: # (# use paws dataset)

[//]: # (dataset_id = 'paws')

[//]: # ()
[//]: # (WORK_DIR = 'workspace')

[//]: # ()
[//]: # (cfg = read_config(model_id))

[//]: # (cfg.train.max_epochs = 2)

[//]: # (cfg.train.work_dir = WORK_DIR)

[//]: # (cfg.train.hooks = cfg.train.hooks = [{)

[//]: # (        'type': 'TextLoggerHook',)

[//]: # (        'interval': 100)

[//]: # (    }])

[//]: # (cfg_file = osp.join(WORK_DIR, 'train_config.json'))

[//]: # (cfg.dump(cfg_file))

[//]: # ()
[//]: # (train_dataset = MsDataset.load(dataset_id, namespace='DAMO_NLP', split='train'))

[//]: # (eval_dataset = MsDataset.load(dataset_id, namespace='DAMO_NLP', split='validation'))

[//]: # ()
[//]: # (kwargs = dict()

[//]: # (    model=model_id,)

[//]: # (    train_dataset=train_dataset,)

[//]: # (    eval_dataset=eval_dataset,)

[//]: # (    cfg_file=cfg_file))

[//]: # ()
[//]: # ()
[//]: # (trainer = build_trainer(default_args=kwargs))

[//]: # ()
[//]: # (print('==============================================================='))

[//]: # (print('pre-trained model loaded, training started:'))

[//]: # (print('==============================================================='))

[//]: # ()
[//]: # (trainer.train())

[//]: # ()
[//]: # (print('==============================================================='))

[//]: # (print('train success.'))

[//]: # (print('==============================================================='))

[//]: # ()
[//]: # (for i in range(cfg.train.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('==============================================================='))

[//]: # (```)

### 模型局限性以及可能的偏差
模型训练数据有限，在特定行业数据上，效果可能存在一定偏差。


## 数据评估及结果
在测试集上的f1为0.915。

### 相关论文以及引用信息
```
@InProceedings{paws2019naacl,
  title = {{PAWS: Paraphrase Adversaries from Word Scrambling}},
  author = {Zhang, Yuan and Baldridge, Jason and He, Luheng},
  booktitle = {Proc. of NAACL},
  year = {2019}
}
```