数据集:

• tner/tweetner7 评估指标:
• f1
• 精确率
• 召回率 模型索引:
• 名称: tner/roberta-large-tweetner7-all 结果:
  • 任务: 名称: 标记分类 类型: token-classification 数据集: 名称: tner/tweetner7 类型: tner/tweetner7 参数: tner/tweetner7 指标:
    • 名称: F1 (test_2021) 类型: f1 值: 0.6574551220340903
    • 名称: 精确率 (test_2021) 类型: precision 值: 0.644212629008989
    • 名称: 召回率 (test_2021) 类型: recall 值: 0.6712534690101758
    • 名称: 宏F1 (test_2021) 类型: f1_macro 值: 0.6124665667529737
    • 名称: 宏精确率 (test_2021) 类型: precision_macro 值: 0.6005167968535563
    • 名称: 宏召回率 (test_2021) 类型: recall_macro 值: 0.625251837701222
    • 名称: 实体跨度F1 (test_2021) 类型: f1_entity_span 值: 0.7881979839166384
    • 名称: 实体跨度精确率 (test_2020) 类型: precision_entity_span 值: 0.7722783264898457
    • 名称: 实体跨度召回率 (test_2021) 类型: recall_entity_span 值: 0.804787787672025
    • 名称: F1 (test_2020) 类型: f1 值: 0.6628787878787878
    • 名称: 精确率 (test_2020) 类型: precision 值: 0.6924816280384398
    • 名称: 召回率 (test_2020) 类型: recall 值: 0.6357031655422937
    • 名称: 宏F1 (test_2020) 类型: f1_macro 值: 0.6297223287745568
    • 名称: 宏精确率 (test_2020) 类型: precision_macro 值: 0.6618492079232416
    • 名称: 宏召回率 (test_2020) 类型: recall_macro 值: 0.601311568050436
    • 名称: 实体跨度F1 (test_2020) 类型: f1_entity_span 值: 0.7642760487144791
    • 名称: 实体跨度精确率 (test_2020) 类型: precision_entity_span 值: 0.7986425339366516
    • 名称: 实体跨度召回率 (test_2020) 类型: recall_entity_span 值: 0.7327451997924235

管道标签: token-classification 小部件:

• 文本: "从 {@herbiehancock@} 通过 {@bluenoterecords@} 获取 Takin' Off 专辑的全模拟经典黑胶版本，链接如下: {{URL}}" 示例标题: "NER 示例 1"

tner/roberta-large-tweetner7-all

此模型是基于 roberta-large 在 tner/tweetner7 数据集（train_all 分割）上微调的版本。模型微调通过 T-NER 的超参数搜索完成（详见仓库）。在2021年的测试集上，它取得了以下结果：

• F1（微平均）: 0.6574551220340903
• 精确率（微平均）: 0.644212629008989
• 召回率（微平均）: 0.6712534690101758
• F1（宏平均）: 0.6124665667529737
• 精确率（宏平均）: 0.6005167968535563
• 召回率（宏平均）: 0.625251837701222

测试集上各实体类别的F1分数如下：

• 公司: 0.5392156862745098
• 创意作品: 0.4760582928521859
• 事件: 0.4673321234119782
• 团体: 0.6139798488664987
• 地点: 0.6707399864222675
• 人物: 0.8293212669683258
• 产品: 0.6906187624750498

F1分数的置信区间通过自助法获得如下：

• F1（微平均）:
  • 90%: [0.6484148010152769, 0.6672289519134409]
  • 95%: [0.6470100684797441, 0.6689850350992637]
• F1（宏平均）:
  • 90%: [0.6484148010152769, 0.6672289519134409]
  • 95%: [0.6470100684797441, 0.6689850350992637]

完整评估结果可在NER指标文件和实体跨度指标文件中找到。

使用方法

此模型可通过tner库使用。通过pip安装库。

pip install tner

TweetNER7 预处理了推文，将账户名和URL转换为特殊格式（详见数据集页面），因此我们需要相应处理推文后运行模型预测，如下所示。

import re
from urlextract import URLExtract
from tner import TransformersNER

extractor = URLExtract()

def format_tweet(tweet):
    # 掩码网页链接
    urls = extractor.find_urls(tweet)
    for url in urls:
        tweet = tweet.replace(url, "{{URL}}")
    # 格式化推特账户
    tweet = re.sub(r"\b(\s*)(@[\S]+)\b", r'\1{\2@}', tweet)
    return tweet


text = "从 @herbiehancock 通过 @bluenoterecords 获取 `Takin' Off` 专辑的全模拟经典黑胶版本，链接如下: http://bluenote.lnk.to/AlbumOfTheWeek"
text_format = format_tweet(text)
model = TransformersNER("tner/roberta-large-tweetner7-all")
model.predict([text_format])

也可以通过transformers库使用，但不推荐，因为目前不支持CRF层。

训练超参数

训练期间使用的超参数如下：

• 数据集: ['tner/tweetner7']
• 数据集分割: train_all
• 数据集名称: 无
• 本地数据集: 无
• 模型: roberta-large
• crf: 是
• 最大长度: 128
• 训练轮数: 30
• 批量大小: 32
• 学习率: 1e-05
• 随机种子: 0
• 梯度累积步数: 1
• 权重衰减: 1e-07
• 学习率预热步比例: 0.15
• 最大梯度范数: 1

完整配置可在微调参数文件中找到。

参考文献

如果使用此模型，请引用T-NER论文和TweetNER7论文。

• T-NER

@inproceedings{ushio-camacho-collados-2021-ner,
    title = "{T}-{NER}: An All-Round Python Library for Transformer-based Named Entity Recognition",
    author = "Ushio, Asahi  and
      Camacho-Collados, Jose",
    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations",
    month = apr,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.eacl-demos.7",
    doi = "10.18653/v1/2021.eacl-demos.7",
    pages = "53--62",
    abstract = "Language model (LM) pretraining has led to consistent improvements in many NLP downstream tasks, including named entity recognition (NER). In this paper, we present T-NER (Transformer-based Named Entity Recognition), a Python library for NER LM finetuning. In addition to its practical utility, T-NER facilitates the study and investigation of the cross-domain and cross-lingual generalization ability of LMs finetuned on NER. Our library also provides a web app where users can get model predictions interactively for arbitrary text, which facilitates qualitative model evaluation for non-expert programmers. We show the potential of the library by compiling nine public NER datasets into a unified format and evaluating the cross-domain and cross- lingual performance across the datasets. The results from our initial experiments show that in-domain performance is generally competitive across datasets. However, cross-domain generalization is challenging even with a large pretrained LM, which has nevertheless capacity to learn domain-specific features if fine- tuned on a combined dataset. To facilitate future research, we also release all our LM checkpoints via the Hugging Face model hub.",
}

• TweetNER7

@inproceedings{ushio-etal-2022-tweet,
    title = "{N}amed {E}ntity {R}ecognition in {T}witter: {A} {D}ataset and {A}nalysis on {S}hort-{T}erm {T}emporal {S}hifts",
    author = "Ushio, Asahi  and
        Neves, Leonardo  and
        Silva, Vitor  and
        Barbieri, Francesco. and
        Camacho-Collados, Jose",
    booktitle = "The 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing",
    month = nov,
    year = "2022",
    address = "Online",
    publisher = "Association for Computational Linguistics",
}