# fiNERweb – Malayalam (mal)

## Dataset Summary

**iNERweb (Malayalam)** is a Malayalam subset of the **fiNERweb** multilingual Named Entity Recognition (NER) dataset.
The dataset provides **character-level entity annotations**, enabling precise span-based NER modeling suitable for transformer and span-classification architectures.

* **Language:** Malayalam (`mal`)
* **Task:** Named Entity Recognition (NER)
* **Annotation type:** Character spans
* **Split:** Train

---

## Dataset Structure

```text
DatasetDict({
    train: Dataset({
        features: ['id', 'text', 'char_spans'],
        num_rows: 27,552
    })
})
```

---

## Data Fields

Each example in the dataset contains the following fields:

```json
{
  "id": str,
  "text": str,
  "char_spans": [
    {
      "start": int,
      "end": int,
      "label": str
    }
  ]
}
```

### Field Description

| Field        | Type         | Description                              |
| ------------ | ------------ | ---------------------------------------- |
| `id`         | `string`     | Unique identifier for the example        |
| `text`       | `string`     | Original Malayalam text                  |
| `char_spans` | `list[dict]` | Character-level entity spans with labels |

---

## Intended Use

This dataset is suitable for:

* Character-span–based NER models
* Transformer-based Malayalam NER (mBERT, XLM-R, IndicBERT, etc.)
* Span classification / token-to-span alignment research
  -д- Multilingual and low-resource NER experiments

---

## Usage

### Load with 🤗 Datasets

```python
from datasets import load_dataset

dataset = load_dataset("Navneeth017/fiNERweb_mal")  # replace with your repo name
print(dataset["train"][0])
```

---

## Source Dataset

This dataset is derived from **fiNERweb**, a multilingual NER dataset containing annotated web text across multiple languages.

Original repository:

```python
load_dataset("whoisjones/fiNERweb", "mal")
```

---

## Citation

If you use this dataset, **please cite the original fiNERweb paper**:

```bibtex
@misc{golde2025finerwebdatasetsartifactsscalable,
  title={FiNERweb: Datasets and Artifacts for Scalable Multilingual Named Entity Recognition},
  author={Jonas Golde and Patrick Haller and Alan Akbik},
  year={2025},
  eprint={2512.13884},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2512.13884}
}
```

---

## Languages

* `mal` — Malayalam

---

## License

Please refer to the **original fiNERweb dataset license** for usage and redistribution terms.