Datasets:
You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This is a commercial dataset. The corpus itself is not hosted on Hugging Face. Access to samples and licensing terms is granted directly by the rights holder, The English Classroom AB (Sweden). Please provide your details and you will be contacted about licensing.
Log in or Sign Up to review the conditions and access this dataset content.
The English Classroom Corpus
A 600-hour, CEFR-graded English language teaching corpus (A1–C2) with aligned native-speaker audio. Single author. Clean rights. Available under commercial licence.
⚠️ The data is not hosted in this repository. This is a dataset card for discovery purposes. Licensing enquiries: the-english-classroom.com/licensing or jennifer@the-englishclassroom.com
Dataset summary
The English Classroom Corpus is a complete English language curriculum authored between 2019 and 2026 by a single qualified UK teacher. It spans all six CEFR levels (A1, A2, B1, B2, C1, C2) and covers all 17 UN Sustainable Development Goals at every level. It was built as one sequenced curriculum rather than an aggregation, so level grading, vocabulary progression and pedagogical scaffolding are consistent across the full 600 hours.
The corpus has been used in real classrooms by 7,000+ teachers across 80+ countries since 2020, including UNRWA-supported education contexts.
What the corpus contains
- 600 hours of classroom-ready lessons across all six CEFR levels
- Teacher instructions, student tasks, and answer keys — natural instruction–response structures
- 2,500+ audio clips recorded by the author (native British English), aligned to lesson content
- CEFR difficulty labels on every text, assigned by a qualified teacher
- SDG topic tags on every lesson (all 17 goals, at every level)
- Source formats: PDF and SCORM; structured text conversion available on request
Why it is distinctive as training data
- Reliable graded difficulty labels for controllable generation, readability modelling, and learner-level adaptation
- Parallel simplification signal: the same topics are written at six proficiency levels, from beginner to near-native — a structure that does not exist in web-scraped data
- Instruction-tuning pairs native to the data: instructions → tasks → answers
- Text–audio alignment from a single speaker for speech and pronunciation applications
- Pedagogical sequencing: lessons are ordered within a real curriculum, providing progression signal
Provenance and rights
- Every lesson and every audio clip authored/recorded by one person: Jennifer Gardner, qualified UK teacher, founder of The English Classroom AB
- No third-party copyrighted material, no user-generated content, no personal data
- All rights held by The English Classroom AB (Sweden), wholly owned by the author
- The corpus is unencumbered: it has never been licensed and sits outside all existing agreements of the rights holder
Licensing
| Licence | Scope |
|---|---|
| AI training licence (non-exclusive) | Training, fine-tuning, evaluation of ML models |
| Product integration licence | Curriculum spine for ed-tech / AI tutoring products, white-label available |
| Institutional licence | Direct classroom use by schools, NGOs, ministries |
Exclusivity considered by sector carve-out only. Sample set available under NDA.
Contact: Jennifer Gardner — jennifer@the-englishclassroom.com — the-english-classroom.com/licensing
Citation
If referencing this dataset card:
The English Classroom Corpus (2026). The English Classroom AB, Kalmar, Sweden.
https://www.the-english-classroom.com/licensing
- Downloads last month
- 12