Lasq Eval

---
name: lasq-eval
description: Probes the ability to extract aspect-based sentiment quadruples (target, aspect, opinion, sentiment) from text in low-resource agglutinative languages. It evaluates exact-match performance across entity detection, relation linking, and full quadruple composition. Use when the user wants to benchmark on LASQ, or asks about evaluating this task. Reports F1.
metadata:
  skill_kind: dataset_eval
  source_arxiv: 2604.10417
  bibtex_key: yusufu2026lasq
  confidence: high
---

# lasq-eval

> LASQ: A Low-resource Aspect-based Sentiment Quadruple Extraction Dataset — Yusufu et al. (arXiv:2604.10417, 2026)

## What this evaluates

Probes the ability to extract aspect-based sentiment quadruples (target, aspect, opinion, sentiment) from text in low-resource agglutinative languages. It evaluates exact-match performance across entity detection, relation linking, and full quadruple composition.

## Datasets

- **LASQ** — total ?; splits: train (-1), dev (-1), test (-1)

## Metrics

- `F1` **(primary)** — range: percent
  - Precision, recall, and F1 are computed based on exact match of predicted quadruples against gold quadruples. A prediction is correct only if all four components (target, aspect, opinion, sentiment) match the gold standard exactly.

## Input / output format

**Input**: Raw text sentence in Uzbek or Uyghur.

**Output**: A list of quadruples, each containing (Target, Aspect, Opinion, Sentiment).

## Scoring recipe

```python
def compute_f1(predictions, golds):
    correct = sum(1 for p in predictions if p in golds)
    precision = correct / len(predictions) if predictions else 0
    recall = correct / len(golds) if golds else 0
    f1 = 2 * precision * recall / (precision + recall) if (precision + recall) > 0 else 0
    return f1 * 100
```

## Common pitfalls

- Exact match is strictly required; partial overlaps or synonym substitutions do not count as correct.
- The task requires extracting four linked components per quadruple, making it sensitive to cascading errors in entity and relation detection.
- Agglutinative morphology in Uzbek and Uyghur can cause tokenization mismatches if subword boundaries are not handled carefully.

## Evidence (verbatim from paper)

> Our evaluation metrics follow [meatwp], using the precision (P), recall (R) and F1. These metrics can be used to detect entities, relations and quadruple. For an item to be considered a correct prediction, it needs to match the gold standard exactly.

## Citation

```bibtex
@misc{yusufu2026lasq,
  title={LASQ: A Low-resource Aspect-based Sentiment Quadruple Extraction Dataset},
  author={Yusufu et al.},
  year={2026},
  note={arXiv:2604.10417}
}
```

- arXiv: 2604.10417