uma-siddareddy/phi2-block-to-text-python-education-dataset

Block-to-Text Python Education Instruction Dataset

Overview

This dataset was created as part of an MSc Computer Science dissertation focused on supporting learners during the transition from block-based programming (such as Blockly) to text-based Python coding.

The dataset is designed for instruction tuning of lightweight language models, particularly Microsoft Phi-2, to act as an educational assistant for beginner programmers rather than a full code generator.

It prioritises clarity, age-appropriate explanations, and minimal guidance to encourage learning and problem-solving.

Educational Context

Block-based programming environments help beginners avoid syntax errors but often fail to prepare learners for real programming languages. When students move to Python, they frequently struggle with:

• —Syntax rules (indentation, colons, brackets)
• —Interpreting error messages
• —Translating visual logic into written code

This dataset directly addresses these challenges by providing structured instructional examples that mirror common beginner mistakes and learning scenarios.

Dataset Format

The dataset is provided in JSONL (JSON Lines) format.

Each line represents one training example using an instruction-tuning structure:

{
  "instruction": "...",
  "input": "...",
  "output": "..."
}