license: cc-by-4.0
tags:
- multilingual
- chatbot
- instructional
- dialogues
- dataset
- nlp
- fine-tuning
- ai-assistant
- huggingface-dataset
- cc-by
- >-
multilingual, instructional, chatbot, dataset, dialogues, assistant,
fine-tuning, cross-lingual, alignment, open-source
Multilingual Instructional Dialogues (10-Language Dataset)
Multilingual Instructional Dialogues is a high-quality dataset of 100 structured, goal-oriented dialogues in 10 major world languages, created for training and fine-tuning AI assistants, chatbots, and instruction-tuned large language models.
Each dialogue simulates a clear, polite interaction where a user asks for guidance on how to perform a task, and the assistant responds with easy-to-follow steps. This dataset has been crafted to maintain consistent conversational structure across languages while respecting the linguistic and cultural nuances of each language.
๐ Languages Included
- ๐บ๐ธ English
- ๐ช๐ธ Spanish
- ๐ซ๐ท French
- ๐ฉ๐ช German
- ๐ง๐ท Portuguese (Brazilian)
- ๐ท๐บ Russian
- ๐ธ๐ฆ Arabic
- ๐ฎ๐ณ Hindi
- ๐จ๐ณ Chinese (Simplified)
- ๐ซ๐ฎ Finnish
๐ง Dataset Highlights
- 100 instructional topics per language
- Polite and instructional dialogue format
- Uniform schema in
.jsonland.csv - Multilingual parity across all languages
- Ready for instruction-tuning and multilingual benchmarking
๐ก Use Cases
- Multilingual chatbot & assistant training
- Fine-tuning instruction-following LLMs
- Dialogue modeling & cross-lingual transfer
- EdTech & AI tutor development
- Language alignment research
๐ File Formats
Each language includes:
instructional_dialogues_full_[LANG].csvinstructional_dialogues_full_[LANG].jsonl
Also included:
instructional_dialogues_multilingual_dataset.zipโ full multilingual packREADME.mdandLICENSE.txt
๐ License
This dataset is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0).
You may use it freely for personal, academic, or commercial purposes, as long as proper credit is given.
https://creativecommons.org/licenses/by/4.0/
๐ฉ Contact
Publisher: Raftico
๐ง toimisto.raftico@gmail.com
We welcome collaboration, feedback, and ideas for future versions or extensions of this dataset.