---
language: en
license: apache-2.0
base_model: bert-base-cased
tags:
  - bert
  - token-classification
  - ner
  - conll2003
datasets:
  - conll2003
metrics:
  - seqeval
pipeline_tag: token-classification
---

# BERT fine-tuned on CoNLL-2003 (NER)

`bert-base-cased` fine-tuned for Named Entity Recognition on [CoNLL-2003](https://huggingface.co/datasets/conll2003).

Recognizes 4 entity types: **PER**, **ORG**, **LOC**, **MISC**.

## Evaluation results

| Metric    | Score  |
|-----------|--------|
| Precision | 0.7058 |
| Recall    | 0.5080 |
| F1        | 0.5908 |
| Accuracy  | 0.9015 |

Evaluated with [seqeval](https://github.com/chakki-works/seqeval) on the CoNLL-2003 test split.

## Usage

```python
from transformers import pipeline

ner = pipeline("ner", model="ZaharHR/bert-conll2003-ner", aggregation_strategy="simple")
ner("Elon Musk founded SpaceX in California.")
```

## Training details

- **Base model:** `bert-base-cased`
- **Dataset:** CoNLL-2003
- **Epochs:** 1
- **Effective batch size:** 16 (gradient accumulation)
- **Optimizer:** AdamW, weight decay 0.01
- **Warmup steps:** 500

## Label scheme

```
O, B-PER, I-PER, B-ORG, I-ORG, B-LOC, I-LOC, B-MISC, I-MISC
```