---
license: mit
tags:
- code
---
# StackOverflow QA Dataset for RAG

## Description

This dataset contains question–answer pairs extracted from the public StackOverflow data dump.

Each example consists of a question with its accepted answer, along with metadata such as tags, scores, and engagement statistics. The dataset is designed for retrieval-augmented generation (RAG), semantic search, and information retrieval tasks.

---

## Dataset Structure

Each row is a JSON object with the following fields:

- `question_id` — unique ID of the question  
- `answer_id` — unique ID of the accepted answer  
- `title` — question title  
- `question_body` — cleaned text of the question  
- `answer_body` — cleaned text of the accepted answer  
- `question_text` — title + question body  
- `combined_text` — title + question + answer (used for retrieval)  
- `tags` — list of StackOverflow tags  
- `question_score` — score of the question  
- `answer_score` — score of the answer  
- `view_count` — number of views  
- `answer_count` — number of answers  
- `favorite_count` — number of favorites  
- `question_creation_date` — timestamp  
- `answer_creation_date` — timestamp