|
--- |
|
license: mit |
|
base_model: dslim/bert-base-NER |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- precision |
|
- recall |
|
- f1 |
|
- accuracy |
|
model-index: |
|
- name: bert-finetuned-ner-ontonotes5 |
|
results: [] |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# Model Description |
|
|
|
This model is a fine-tuned version of [dslim/bert-base-NER](https://huggingface.co/dslim/bert-base-NER) on OntoNotes 5 dataset and is designed to identify and classify various types of entities in text, including persons, organizations, locations, dates, and more. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.1634 |
|
- Precision: 0.8620 |
|
- Recall: 0.8849 |
|
- F1: 0.8733 |
|
- Accuracy: 0.9758 |
|
|
|
## Intended uses & limitations |
|
|
|
The model is intended for use in applications requiring NER, such as information extraction, text classification, and enhancing search capabilities by identifying key entities within the text. It can be used to identify entities in any English text, including news articles, social media posts, and legal documents. |
|
|
|
|
|
## Training and evaluation data |
|
|
|
Training Data |
|
The model was fine-tuned on the OntoNotes 5 dataset. This dataset includes multiple types of named entities and is widely used for NER tasks. The dataset is annotated with the following entity tags: |
|
|
|
CARDINAL: Numerical values |
|
|
|
DATE: References to dates and periods |
|
|
|
PERSON: Names of people |
|
|
|
NORP: Nationalities, religious groups, political groups |
|
|
|
GPE: Countries, cities, states |
|
|
|
LAW: Named documents and legal entities |
|
|
|
ORG: Organizations |
|
|
|
PERCENT: Percentage values |
|
|
|
ORDINAL: Ordinal numbers |
|
|
|
MONEY: Monetary values |
|
|
|
WORK_OF_ART: Titles of creative works |
|
|
|
FAC: Facilities |
|
|
|
TIME: Times smaller than a day |
|
|
|
LOC: Non-GPE locations, mountain ranges, bodies of water |
|
|
|
QUANTITY: Measurements, as of weight or distance |
|
|
|
PRODUCT: Objects, vehicles, foods, etc. (not services) |
|
|
|
EVENT: Named events |
|
|
|
LANGUAGE: Named languages |
|
|
|
## Model Configuration |
|
Base Model: dslim/bert-base-NER |
|
|
|
Number of Labels: 37 (including the "O" tag for outside any named entity) |
|
|
|
### 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: 5 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | |
|
|:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:| |
|
| 0.0372 | 1.0 | 7491 | 0.1188 | 0.8392 | 0.8799 | 0.8591 | 0.9738 | |
|
| 0.04 | 2.0 | 14982 | 0.1182 | 0.8562 | 0.8824 | 0.8691 | 0.9754 | |
|
| 0.0164 | 3.0 | 22473 | 0.1380 | 0.8561 | 0.8835 | 0.8696 | 0.9752 | |
|
| 0.0117 | 4.0 | 29964 | 0.1531 | 0.8618 | 0.8833 | 0.8724 | 0.9758 | |
|
| 0.0054 | 5.0 | 37455 | 0.1634 | 0.8620 | 0.8849 | 0.8733 | 0.9758 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.41.2 |
|
- Pytorch 2.3.0+cu121 |
|
- Datasets 2.20.0 |
|
- Tokenizers 0.19.1 |
|
|
|
## Contact Information |
|
|
|
For questions, comments, or issues with the model, please contact: |
|
|
|
Name: [Irechukwu Nkweke] |
|
|
|
Email: [mnkweke@yahoo.com] |
|
|
|
GitHub: [https://github.com/mnkweke] |
|
|
|
## Acknowledgments |
|
|
|
This model was trained using the Hugging Face transformers library and the OntoNotes 5 dataset. |
|
|
