hiennthp/esg-bank-dataset-v4

---
language:
- vi
- en
license: cc-by-nc-4.0
task_categories:
- text-classification
tags:
- esg
- greenwashing-detection
- banking
- vietnam
- sustainability
size_categories:
- 1K<n<10K
---

# Vietnamese Banking ESG Disclosure Dataset

## Description

Multi-task annotated dataset for ESG (Environmental, Social, Governance)
disclosure analysis in the Vietnamese banking sector. Each sample is a
text chunk from bank sustainability/annual reports, annotated with:

- **Greenwashing classification** (Legitimate / Greenwashing / Uncertain)
- **ESG pillar** (Environmental / Social / Governance / General)
- **Content quality score** (0-100)
- **ESG score** (0-100)

## Dataset Statistics

| Split | Samples |
|-------|---------|
| train | 2,059 |
| val | 558 |
| test | 900 |
| **Total** | **3,517** |

## Schema

| Field | Type | Description |
|-------|------|-------------|
| text | string | ESG disclosure text chunk |
| greenwash_label | string | Legitimate / Greenwashing / Uncertain |
| pillar_label | string | Environmental / Social / Governance / General |
| quality_score | float | Content quality score (0-100) |
| esg_score | float | Predicted ESG performance score (0-100) |
| report_type | string | SUSTAINABILITY or ANNUAL |
| report_year | int | Report fiscal year |
| bank_id | string | Anonymized bank identifier |
| assurance_level | string | Third-party assurance level |
| has_scope3 | bool | Contains Scope 3 emission data |
| gri_compliant | bool | GRI Standards compliant |
| tcfd_elements_count | int | TCFD recommendation coverage (0-11) |
| circular17_compliant | bool | SBV Circular 17 compliant |

## Annotation Process

1. Initial labeling via LLM (Step 2) with rule-based fallback
2. Content quality scoring (Step 3)
3. Human review and validation (Step 4)

## Citation

If you use this dataset, please cite:
```
[Citation to be added upon publication]
```