---
language:
- ur
task_categories:
- text-classification
tags:
- roman-urdu
- event-detection
- nlp
- urdu
- pakistan
license: cc-by-4.0
size_categories:
- 1K<n<10K
---

# LORU-ED: Labeled Original Roman Urdu Dataset for Event Detection

## Dataset Description
LORU-ED is a novel Roman Urdu dataset for event detection, containing 7,694 
annotated sentences balanced equally between event and non-event classes.

## Dataset Structure
| Split | Size |
|-------|------|
| Train | 6,155 |
| Test  | 1,539 |

## Features
- `id`: Unique identifier
- `text`: Roman Urdu sentence
- `language`: Language tag (roman_urdu/mixed/en)
- `is_event`: Binary label (1=event, 0=non-event)
- `event_type`: Fine-grained event category (17 types)
- `time_raw`: Temporal expression if present
- `negation`: Negation flag
- `past_tense`: Tense flag

## Event Types
exam, event, class, travel, prayer, sports, meeting, financial, 
medical, appointment, government, interview, utility, assignment, 
food, online_event, family

## Citation
If you use this dataset, please cite:
```
@dataset{loru-ed-2026,
  author    = {Adeel Shehbaz},
  title     = {LORU-ED: Roman Urdu Dataset for Event Detection},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/Adeel1percentdotcom/LORU-ED_Roman-Urdu-Event-Detection}
}
```