language: "en" tags:

• twitter
• 立场检测
• 2020美国大选
• 政治 license: "gpl-3.0"

针对乔·拜登立场检测的2020年美国大选推特预训练BERT模型（KE-MLM）

本模型为《立场检测的知识增强掩码语言模型》（NAACL 2021）中KE-MLM模型的预训练权重。

训练数据

该模型基于超过500万条关于2020年美国总统大选的英文推文进行预训练，随后使用我们标注立场的数据集针对乔·拜登的立场检测任务进行微调。

训练目标

该模型以BERT-base为基础，通过标准掩码语言模型目标进行训练，并针对乔·拜登的立场检测任务微调分类层。

使用说明

本预训练语言模型专为针对乔·拜登的立场检测任务微调。

更多细节请参阅官方代码库。

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import numpy as np

# 优先使用GPU
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# 选择模型路径
pretrained_LM_path = "kornosk/bert-election2020-twitter-stance-biden-KE-MLM"

# 加载模型
tokenizer = AutoTokenizer.from_pretrained(pretrained_LM_path)
model = AutoModelForSequenceClassification.from_pretrained(pretrained_LM_path)

id2label = {
    0: "反对",
    1: "支持",
    2: "中立"
}

##### 预测中立立场 #####
sentence = "Hello World."
inputs = tokenizer(sentence.lower(), return_tensors="pt")
outputs = model(**inputs)
predicted_probability = torch.softmax(outputs[0], dim=1)[0].tolist()

print("文本:", sentence)
print("预测结果:", id2label[np.argmax(predicted_probability)])
print("反对概率:", predicted_probability[0])
print("支持概率:", predicted_probability[1])
print("中立概率:", predicted_probability[2])

##### 预测支持立场 #####
sentence = "Go Go Biden!!!"
inputs = tokenizer(sentence.lower(), return_tensors="pt")
outputs = model(**inputs)
predicted_probability = torch.softmax(outputs[0], dim=1)[0].tolist()

print("文本:", sentence)
print("预测结果:", id2label[np.argmax(predicted_probability)])
print("反对概率:", predicted_probability[0])
print("支持概率:", predicted_probability[1])
print("中立概率:", predicted_probability[2])

##### 预测反对立场 #####
sentence = "Biden is the worst."
inputs = tokenizer(sentence.lower(), return_tensors="pt")
outputs = model(**inputs)
predicted_probability = torch.softmax(outputs[0], dim=1)[0].tolist()

print("文本:", sentence)
print("预测结果:", id2label[np.argmax(predicted_probability)])
print("反对概率:", predicted_probability[0])
print("支持概率:", predicted_probability[1])
print("中立概率:", predicted_probability[2])

# 如果觉得有用，请考虑引用我们的论文 :)

参考文献

• 《立场检测的知识增强掩码语言模型》, NAACL 2021.

引用

@inproceedings{kawintiranon2021knowledge,
    title={Knowledge Enhanced Masked Language Model for Stance Detection},
    author={Kawintiranon, Kornraphop and Singh, Lisa},
    booktitle={Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
    year={2021},
    publisher={Association for Computational Linguistics},
    url={https://www.aclweb.org/anthology/2021.naacl-main.376}
}