err0rgod/india_history_semantic_search

---

license: mit task_categories: - text-retrieval - question-answering language: - en tags: - history - india - semantic-search - rag - tutorial size_categories: - n<1K

Indian History Dataset for Semantic Search 🇮🇳

Dataset Summary

This dataset contains 15 detailed Markdown (.md) documents covering major eras, empires, and events in the history of India. It was specifically generated to serve as a practice dataset for learning Semantic Search, Vector Embeddings, and Retrieval-Augmented Generation (RAG).

Each document is structured into distinct thematic sections (Overview, Key Events, and Legacy) to allow practitioners to experiment with text chunking strategies (such as MarkdownHeaderTextSplitter or RecursiveCharacterTextSplitter).

Supported Tasks

• Semantic Search: Practice generating embeddings and retrieving the most semantically relevant historical chunks based on user queries.
• Retrieval-Augmented Generation (RAG): Use the retrieved chunks as context for an LLM to answer historical questions accurately.
• Text Chunking Practice: Experiment with different chunk sizes and overlaps using the consistent Markdown headers (#, ##).

Dataset Structure

The dataset consists of 15 plaintext Markdown files. Average document length: ~150-200 words.

Files included:

1. 01_indus_valley.md
2. 02_vedic_period.md
3. 03_maurya_empire.md
4. 04_gupta_empire.md
5. 05_delhi_sultanate.md
6. 06_mughal_empire.md
7. 07_maratha_empire.md
8. 08_east_india_company.md
9. 09_rebellion_of_1857.md
10. 10_indian_national_congress.md
11. 11_mahatma_gandhi.md
12. 12_partition_and_independence.md
13. 13_integration_of_states.md
14. 14_constitution_of_india.md
15. 15_economic_liberalization.md

Example Use Case (Python/LangChain)

from langchain.text_splitter import MarkdownHeaderTextSplitter

headers_to_split_on = [
    ("#", "Title"),
    ("##", "Section")
]

markdown_splitter = MarkdownHeaderTextSplitter(headers_to_split_on=headers_to_split_on)
# Pass the content of the markdown files here to chunk them perfectly by section!

  ### Step 3: Save
  Scroll down to the bottom of the page and click **"Commit new file"**.

  As soon as you commit, Hugging Face will read the `---` tags at the top and automatically add the tags (like `text-retrieval`, `english`) to your
dataset's page, and the formatted text will appear beautifully on the main landing page!