metadata

datasets:
  - conll2003
metrics:
  - f1
  - precision
  - recall
model-index:
  - name: tner/deberta-v3-large-conll2003
    results:
      - task:
          name: Token Classification
          type: token-classification
        dataset:
          name: conll2003
          type: conll2003
          args: conll2003
        metrics:
          - name: F1
            type: f1
            value: 0.9222388190844389
          - name: Precision
            type: precision
            value: 0.9154020582592011
          - name: Recall
            type: recall
            value: 0.9291784702549575
          - name: F1 (macro)
            type: f1_macro
            value: 0.9043961692086329
          - name: Precision (macro)
            type: precision_macro
            value: 0.8959854326377331
          - name: Recall (macro)
            type: recall_macro
            value: 0.9135442454672595
          - name: F1 (entity span)
            type: f1_entity_span
            value: 0.960570322126386
          - name: Precision (entity span)
            type: precision_entity_span
            value: 0.9550227511375569
          - name: Recall (entity span)
            type: recall_entity_span
            value: 0.9661827195467422
pipeline_tag: token-classification
widget:
  - text: Jacob Collier is a Grammy awarded artist from England.
    example_title: NER Example 1

tner/deberta-v3-large-conll2003

This model is a fine-tuned version of microsoft/deberta-v3-large on the tner/conll2003 dataset. Model fine-tuning is done via T-NER's hyper-parameter search (see the repository for more detail). It achieves the following results on the test set:

F1 (micro): 0.9222388190844389
Precision (micro): 0.9154020582592011
Recall (micro): 0.9291784702549575
F1 (macro): 0.9043961692086329
Precision (macro): 0.8959854326377331
Recall (macro): 0.9135442454672595

The per-entity breakdown of the F1 score on the test set are below:

location: 0.9407496977025392
organization: 0.9115486335586247
other: 0.7920110192837466
person: 0.9732753262896209

For F1 scores, the confidence interval is obtained by bootstrap as below:

F1 (micro):
- 90%: [0.9157944386463721, 0.9286928993636353]
- 95%: [0.9146558483630953, 0.9297919809412201]
F1 (macro):
- 90%: [0.9157944386463721, 0.9286928993636353]
- 95%: [0.9146558483630953, 0.9297919809412201]

Full evaluation can be found at metric file of NER and metric file of entity span.

Training hyperparameters

The following hyperparameters were used during training:

dataset: ['tner/conll2003']
dataset_split: train
dataset_name: None
local_dataset: None
model: microsoft/deberta-v3-large
crf: False
max_length: 128
epoch: 15
batch_size: 16
lr: 1e-05
random_seed: 42
gradient_accumulation_steps: 4
weight_decay: None
lr_warmup_step_ratio: 0.1
max_grad_norm: 10.0

The full configuration can be found at fine-tuning parameter file.

Reference

If you use any resource from T-NER, please consider to cite our paper.


@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.",
}