https://github.com/valenradovich/named-entity-recognition
fine-tuning BERT to achieve token classification, in this case, NER.
https://github.com/valenradovich/named-entity-recognition
bert bert-fine-tuning named-entity-recognition ner pytorch transformers
Last synced: 3 months ago
JSON representation
fine-tuning BERT to achieve token classification, in this case, NER.
- Host: GitHub
- URL: https://github.com/valenradovich/named-entity-recognition
- Owner: valenradovich
- Created: 2024-06-19T17:52:39.000Z (12 months ago)
- Default Branch: master
- Last Pushed: 2024-06-19T17:57:38.000Z (12 months ago)
- Last Synced: 2025-01-20T01:26:07.487Z (5 months ago)
- Topics: bert, bert-fine-tuning, named-entity-recognition, ner, pytorch, transformers
- Language: Jupyter Notebook
- Homepage:
- Size: 118 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# named entity recognition: fine-tuning BERT
fine-tune BERT model for named entity recognition (NER), also called entity extraction, using pytorch and transformers.
you'll find the code in the .ipynb format, which you can run in google colab or jupyter notebook and the .py format (found the example on hugging face transformer documentation), which you can run in your local machine or VM.
the notebook is just easier to run and understand (in colab is ready-to-go), but the .py file is suitable to use as script.
both are very organized and can be reproduced with any dataset just following the instructions in the code.
[CONLL 2003 dataset](https://aclanthology.org/W03-0419.pdf) is used in this example.
## results
the model was trained for 3 epochs and achieved the following results in colab:
```
{'eval_loss': 0.06109246239066124,
'eval_precision': 0.9292045202747617,
'eval_recall': 0.9382481261886118,
'eval_f1': 0.933704425271361,
'eval_accuracy': 0.9837958917819754,
'eval_runtime': 5.937,
'eval_samples_per_second': 547.411,
'eval_steps_per_second': 34.361,
'epoch': 3.0}
```
same with the script:
## how to run
the notebook is ready to run in google colab, just open it and run all cells.
the script can be run in your local machine or VM, just run the following command in the terminal to train the model in the same dataset as i did:
```bash
run.sh
```
## requirements
```bash
pip install -r requirements.txt
```