EsferSami/Bangla-Mental-Health-Dataset-V2

---
language: bn
license: cc-by-4.0
tags:
- bangla
- mental-health
- nlp
- synthetic
- alpaca-format
- dataset
dataset_info: 10,000 synthetic Bangla mental health conversations in input‑instruction‑output
format. Generated from books, newspapers, YouTube shows, blogs and articles. Available
in JSONL, CSV, Parquet and TOON.
---
# Bangla Mental Health Dataset V2.0

## 📘 Overview
This dataset contains **10,000 synthetic Bangla mental health conversations**, designed for:
- LLM fine-tuning
- Mental health classification
- Mental-health dialogue modeling
- Research on Bangla NLP

The dataset includes **input–instruction–output** style records suitable for:
- Alpaca format
- SFT (Supervised Finetuning)
- QLoRA, LoRA
- Chat-style fine-tuning
- 40 Mental helth condition(eg. মানসিক চাপ,আত্মহত্যা,আত্মবিশ্বাস,ভালোবাসাহীনতা,অতিরিক্ত চিন্তা....)

---

## 📚 Source Methodology

This dataset is **synthetically generated**, created using:
- Research from multiple mental-health books
- Newspaper topics and psychological reports
- YouTube talk shows about mental health
- Articles, blogs, and clinical guidance content

> **No real user data or personal information is included.**
> Everything is **synthetic** and constructed through text generation + controlled transformations.

---

## 📂 Provided Formats

This dataset repo provides **four different formats** so users can choose what works best:

| Format | Filename | Purpose |
|--------|----------|----------|
| JSONL | `data/dataset.jsonl` | Best for LLM finetuning (Alpaca, SFT, QLoRA) |
| CSV | `data/dataset.csv` | Good for analysis or simple loading |
| Parquet | `data/dataset.parquet` | Efficient, column-optimized format |
| TOON | `data/dataset.toon` | Token-optimized binary format for fast LLM training |

---

## 🔧 Loading Examples

### Load JSONL,TOON,CSV
```python
from datasets import load_dataset
df = load_dataset("EsferSami/Bangla-Mental-Health-Dataset-V2", data_files="data/dataset.jsonl")

### Load TOON
df = load_dataset("EsferSami/Bangla-Mental-Health-Dataset-V2", data_files="data/dataset.toon")

### Load CSV
df = load_dataset("EsferSami/Bangla-Mental-Health-Dataset-V2", data_files="data/dataset.csv")