README.md

metadata

license: apache-2.0
task_categories:
  - text-generation
language:
  - en
tags:
  - code
pretty_name: 50K Stack Overflow Q&A Dataset
size_categories:
  - 10K<n<100K

50K Stack Overflow Q&A Dataset

A curated collection of 50,000 high-quality Stack Overflow question-answer pairs, filtered for quality and diversity across multiple programming languages.

Dataset Overview

Metric	Value
Total Q&A Pairs	50,000
Average Question Score	267.16
Average Answer Score	339.50
Highest Question Score	26,621
Minimum Question Score	93

Available Formats

Format	File	Size
CSV	50K_stackoverflow.csv	111.60 MB
Parquet	50K_stackoverflow.parquet	58.37 MB
JSONL	50K_stackoverflow.jsonl	125.24 MB

Programming Language Distribution

The dataset covers questions across multiple programming languages and technologies:

Language	Questions	Percentage
JavaScript	333	24.1%
Python	273	19.8%
Java	132	9.6%
PHP	102	7.4%
TypeScript	100	7.2%
C#	91	6.6%
Go	82	5.9%
Ruby	73	5.3%
C++	63	4.6%
Swift	28	2.0%
SQL	17	1.2%
Other	706	6.3%

Dataset Schema

Each record contains the following fields:

Field	Type	Description
question_title	string	The title of the Stack Overflow question
question_body	string	Full text content of the question
question_score	integer	Community upvotes on the question
answer_body	string	The accepted/top-voted answer text
answer_score	integer	Community upvotes on the answer
tags	string	Pipe-separated tags (e.g., java|c++|performance)

Sample Entry

Question Title: Why is processing a sorted array faster than processing an unsorted array?

Tags: java | c++ | performance | cpu-architecture | branch-prediction

Metric	Score
Question Score	26,621
Answer Score	34,269

Quality Filtering Criteria

This dataset was curated with the following quality filters:

• Minimum question score threshold of 93
• Only includes questions with accepted or highly-voted answers
• Covers diverse programming topics and languages
• Excludes closed, duplicate, or low-quality posts

Use Cases

This dataset is ideal for:

• LLM Fine-tuning: Train models to answer programming questions
• Code Understanding: Build systems that comprehend code-related queries
• Information Retrieval: Develop semantic search for technical documentation
• Question Answering Systems: Create domain-specific QA models
• Educational Tools: Power programming tutors and learning assistants
• Text Classification: Train models to categorize technical content

Loading the Dataset

Python (Pandas)

import pandas as pd

# Load CSV
df = pd.read_csv('50K_stackoverflow.csv')

# Load Parquet (recommended for performance)
df = pd.read_parquet('50K_stackoverflow.parquet')

Python (JSONL)

import json

data = []
with open('50K_stackoverflow.jsonl', 'r') as f:
    for line in f:
        data.append(json.loads(line))

Hugging Face Datasets

from datasets import load_dataset

dataset = load_dataset('Omarrran/50K_stackoverflow_qna_dataset_by_hnm')

File Format Recommendations

Use Case	Recommended Format
Data Analysis	Parquet
Machine Learning	Parquet
Streaming/Processing	JSONL
Spreadsheet Tools	CSV
Storage Efficiency	Parquet (52% smaller than CSV)

Citation

If you use this dataset in your research or projects, please cite:

@dataset{stackoverflow_50k_qa,
  title={Omarrran/50K_stackoverflow_qna_dataset_by_hnma},
  Author= {Haq Nawaz Malik}
  year={2025},
  url={https://huggingface.co/datasets/Omarrran/50K_stackoverflow_qna_dataset_by_hnm/}
  
}

License

This dataset is derived from Stack Overflow content, which is licensed under CC BY-SA 4.0. Any use of this dataset must comply with Stack Overflow's terms of service and attribution requirements.

Acknowledgments

• Data sourced from Stack Overflow

---

Note: This dataset represents a snapshot of Stack Overflow content and may not reflect the most current answers or best practices for rapidly evolving technologies.