# License This corpus is a **mixed-source dataset** with multiple licenses. Each row is tagged with a `license` field indicating its source license. ## License Breakdown by Source ### IRIISNEPAL Dataset - **License:** MIT - **Rows:** ~6.0 million - **Source:** https://github.com/bnltm/IIIiSNepal - **Applies to:** `nepali-formal-corpus` ### Nepali Wikipedia - **License:** CC BY-SA 4.0 - **Rows:** ~291,000 - **Source:** Nepali Wikipedia dump - **Attribution:** Derived from Wikipedia contributors - **Applies to:** `nepali-formal-corpus` ### Scraped News Sources - **License:** Source-dependent - **Rows:** ~87,000 - **Sources:** - Kantipur News (private use) - Setopati (private use) - Nepal Khabar (private use) - Nagarik (private use) - Other Nepali news outlets - **Note:** News articles are included under fair-use research assumptions. Please verify licensing with original publishers for commercial use. - **Applies to:** `nepali-formal-corpus` ### YouTube Comments - **License:** CC BY 4.0 - **Rows:** ~431,000 - **Source:** YouTube API (public comments) - **Attribution:** YouTube commenters - **Applies to:** `nepali-colloquial-corpus`, `roman-nepali-corpus` ## Usage Guidance - **For academic research:** All sources are permissible under academic fair-use principles. - **For commercial use:** Filter by license field. Use only MIT and CC BY 4.0 rows, or obtain permissions from news outlets. - **For compliance:** Review the `license` column in each dataset to filter by acceptable licenses for your use case. ## Recommended Attribution If you use this corpus, cite it as: ``` @dataset{nepali_text_corpus_2026, author = {Aadarsha Chhetri}, title = {Nepali Text Corpus}, year = {2026}, organization = {Hugging Face Datasets}, url = {https://huggingface.co/datasets/Boredoom17/nepali-text-corpus} } ``` And acknowledge individual sources: - IRIISNEPAL for formal text - Wikipedia contributors for encyclopedia content - Nepali news outlets for journalism - YouTube commenters for colloquial speech --- **Last Updated:** April 2, 2026