bert-finetuned-ner

This model is a fine-tuned version of bert-base-cased on the wnut_17 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4652
• Precision: 0.5994
• Recall: 0.4797
• F1: 0.5329
• Accuracy: 0.9245

Model description

bert-finetuned-ner is a fine-tuned BERT model aimed at performing Named Entity Recognition (NER) tasks. This model is particularly fine-tuned on the WNUT-17 dataset, which includes a variety of unusual and emerging named entities that are difficult for traditional NER systems to recognize

Intended uses & limitations

Intended uses

Named Entity Recognition (NER) for identifying unusual and emerging entities Use cases in social media text, conversational agents, and user-generated content where new and rare entities frequently appear

Limitations

The model may not perform well on datasets significantly different from WNUT-17 It might struggle with very domain-specific entities not covered during training

Training and evaluation data

The model was trained and evaluated on the WNUT-17 dataset. This dataset is specifically designed to test models on their ability to recognize emerging and rare named entities in noisy text data.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	425	0.4480	0.5579	0.4498	0.4980	0.9229
0.0345	2.0	850	0.4335	0.5589	0.4653	0.5078	0.9235
0.0325	3.0	1275	0.4652	0.5994	0.4797	0.5329	0.9245

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1