Somtharu181coder's picture
Update README.md
e9b1b11 verified
|
Raw
History Blame Contribute Delete
2.1 kB

Nepali Education SFT Dataset

A high-quality Nepali-language Supervised Fine-Tuning (SFT) dataset designed for instruction-following and educational question-answering tasks.

The dataset is structured in a conversational format and contains Nepali educational instructions and responses along with detailed source and generation metadata. It was developed with emphasis on data integrity, language purity, structural consistency, intent coverage, and dataset quality assurance.


📌 Dataset Overview

Property Value
Dataset Records 1999
Columns 19
Language Nepali
Script Devanagari
Task Type SFT / Instruction Following
Data Format JSONL
Conversation Format Yes
Malformed Records 0
Incomplete Records 0
Duplicate IDs 0
Duplicate Complete Rows 0
Duplicate Questions 0
Near-Duplicate Question Pairs 0
Strong Near-Duplicate Pairs 0
Intent Coverage 100%
Rows Requiring Review 0
Corrupted Rows 0
Reported Overall Quality Score 100%

🎯 Purpose

The primary purpose of this dataset is to provide high-quality Nepali educational instruction-response examples suitable for Supervised Fine-Tuning of Large Language Models.

The dataset is designed to support models in learning:

  • Nepali instruction following
  • Educational question answering
  • Conversational interaction
  • Intent recognition
  • Nepali-language response generation
  • Educational knowledge delivery
  • Different question-answering behaviors

The dataset also contains metadata that can be used for dataset analysis, filtering, auditing, and future quality-control workflows.


🧠 Dataset Design

The dataset follows a structured conversational SFT format.

Each example contains a conversations field representing the interaction between the user and the assistant.

Conceptually, an example follows the structure:

{
  "id": "1",
  "conversations": [
    {
      "role": "user",
      "content": "..."
    },
    {
      "role": "assistant",
      "content": "..."
    }
  ]
}