README.md

metadata

license: mit
dataset_info:
  features:
    - name: title
      dtype: large_string
    - name: video_id
      dtype: large_string
    - name: transcript
      dtype: large_string
  splits:
    - name: train
      num_bytes: 130792887
      num_examples: 1192
  download_size: 61288449
  dataset_size: 130792887
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - question-answering
  - text-generation
language:
  - en
tags:
  - code
pretty_name: Free Code Camp Transcripts
size_categories:
  - 1K<n<10K

Free Code Camp Transcripts

Overview

This dataset contains transcripts of programming tutorials from FreeCodeCamp videos. Each entry includes the video title, YouTube video ID, and the full transcript, making it suitable for training and evaluating NLP and LLM systems focused on developer education.

DataSource

---

Dataset Structure

Column	Type	Description
title	string	Title of the YouTube video
video_id	string	Unique YouTube video identifier
transcript	string	Full transcript of the video

---

Dataset Details

• Total Samples: 1,192
• Language: English
• Format: Parquet (auto-converted by Hugging Face)
• Domain: Programming / Software Development

---

How to Load the Dataset

from datasets import load_dataset

dataset = load_dataset("nuhmanpk/freecodecamp-transcripts")
print(dataset)

print(dataset["train"][0])

---

Example Record

{
  "title": "PostgreSQL Tutorial for Beginners",
  "video_id": "SpfIwlAYaKk",
  "transcript": "Welcome to this PostgreSQL tutorial..."
}

---

Use Cases

1. Text Summarization

from transformers import pipeline

summarizer = pipeline("summarization")

text = dataset["train"][0]["transcript"]
summary = summarizer(text[:2000])

print(summary)

---

2. Question Answering

from transformers import pipeline

qa = pipeline("question-answering")

context = dataset["train"][0]["transcript"]
question = "What is PostgreSQL?"

result = qa(question=question, context=context)
print(result)

---

3. Instruction Dataset

def to_instruction(example):
    return {
        "prompt": f"Explain this tutorial: {example['title']}",
        "response": example["transcript"][:1000]
    }

instruction_ds = dataset["train"].map(to_instruction)

---

4. Embeddings

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("all-MiniLM-L6-v2")

embeddings = model.encode(dataset["train"]["transcript"][:100])

---

Preprocessing Tips

dataset = dataset.filter(lambda x: x["transcript"] != "")

def chunk_text(text, size=1000):
    return [text[i:i+size] for i in range(0, len(text), size)]

---

Limitations

• Transcripts may contain noise
• No timestamps
• Limited to programming tutorials

---

License

MIT License

---

Future Improvements

• Add topic tags
• Generate QA pairs
• Instruction tuning

---