README.md

metadata

dataset_info:
  features:
    - name: conversation_id
      dtype: int64
    - name: user
      dtype: string
    - name: assistant
      dtype: string
  splits:
    - name: train
      num_bytes: 2927762
      num_examples: 15187
  download_size: 123936
  dataset_size: 2927762
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
task_categories:
  - question-answering
language:
  - bn
pretty_name: AI conversation

Bengali Multi-turn Chat Dataset

A high-quality Bengali conversational dataset derived from chatbot interactions.
Each row contains a single turn of conversation with:

• conversation_id — identifies a continuous multi-turn dialogue
• user — Bengali user message
• assistant — chatbot response in Bengali

---

📂 Dataset Summary

This dataset contains multi-turn conversational pairs in Bengali, designed for developing and fine-tuning conversational AI models such as GPT-2, GPT-Neo or other dialogue models.

✔ Fully Bengali
✔ Natural multi-turn chat style
✔ Good for instruction-style chatbots
✔ Clean user-assistant formatting

---

📊 Dataset Structure

Field	Type	Description
conversation_id	int	Group of dialogue turns
user	string	Bengali user input
assistant	string	Bengali model output

Example:

{
  "conversation_id": 2,
  "user": "তোমার লক্ষ্য কী?",
  "assistant": "আমার লক্ষ্য হলো তথ্য সহজলভ্য করা।"
}