|
|
--- |
|
|
license: cc-by-4.0 |
|
|
language: ar |
|
|
tags: |
|
|
- named-entity-recognition |
|
|
- arabic |
|
|
- human-annotated |
|
|
- iahlt |
|
|
dataset_name: iahlt-ner-arabic |
|
|
--- |
|
|
|
|
|
# IAHLT Named Entities Dataset (Arabic Subset) |
|
|
|
|
|
**האיגוד הישראלי לטכנולוגיות שפת אנוש** |
|
|
**الرابطة الإسرائيلية لتكنولوجيا اللغة البشرية** |
|
|
**The Israeli Association of Human Language Technologies** |
|
|
https://www.iahlt.org |
|
|
|
|
|
This dataset contains named entity annotations for Arabic texts from various sources, curated as part of the IAHLT multilingual NER project. The Arabic portion is provided here as a cleaned subset intended for training and evaluation in named entity recognition tasks. |
|
|
|
|
|
## Files Included |
|
|
|
|
|
This release includes the following JSONL files: |
|
|
|
|
|
- `iahlt_ner_train.jsonl` |
|
|
- `iahlt_ner_val.jsonl` |
|
|
- `iahlt_ner_test.jsonl` |
|
|
|
|
|
Each file contains one JSON object per line with the following fields: |
|
|
- `text`: the raw paragraph text |
|
|
- `label`: a list of triples `[start, end, label]` where: |
|
|
- `start` is the index of the first character of the entity |
|
|
- `end` is the index of the first character after the entity |
|
|
- `label` is the entity class (e.g., `PER`, `ORG`, etc.) |
|
|
- `metadata`: a dictionary with source and annotation metadata |
|
|
|
|
|
## Entity Types |
|
|
|
|
|
The dataset includes the following entity types (summed across splits): |
|
|
|
|
|
| Entity | Count | |
|
|
|-----------|-------| |
|
|
| GPE | 26,767 | |
|
|
| PER | 22,694 | |
|
|
| ORG | 19,906 | |
|
|
| TIMEX | 10,288 | |
|
|
| TTL | 10,075 | |
|
|
| FAC | 4,740 | |
|
|
| MISC | 7,087 | |
|
|
| LOC | 5,389 | |
|
|
| EVE | 2,595 | |
|
|
| WOA | 1,781 | |
|
|
| DUC | 1,715 | |
|
|
| ANG | 732 | |
|
|
| INFORMAL | 10 | |
|
|
|
|
|
> Note: These statistics reflect a cleaned version of the dataset. Some entities and texts have been modified or removed for consistency and usability. |
|
|
|
|
|
## Annotation Notes |
|
|
|
|
|
- Entity spans were manually annotated at the grapheme level and then normalized using Arabic-specific punctuation and spacing rules. |
|
|
- Nested spans are allowed. |
|
|
- The dataset was cleaned to ensure compatibility with common NER training formats. |
|
|
|
|
|
## License |
|
|
|
|
|
This dataset is released under the [Creative Commons Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/) license. |
|
|
|
|
|
## Acknowledgments |
|
|
|
|
|
We thank all annotators and contributors who worked on this corpus. |
|
|
|
