The Louie Project Sentence Bank: a graded ten-language parallel corpus of English example sentences for children

A graded, ten-language parallel corpus of English example sentences written for children learning English outside a classroom. 13,133 sentences and 647 vocabulary items, each aligned with glosses in Korean, Japanese, Chinese, Vietnamese, Spanish, Portuguese, German, Indonesian and French, and each labelled with the school level and topic collection it belongs to.

Scope and structure. Two independent level ladders are represented: the United States school sequence (Kindergarten to Grade 9) and the Korean school sequence (elementary 1-6, middle 1-3). Collections that are not level-bound - everyday conversation, books and media, video games - are marked general. Sentences are either single statements or turns inside multi-turn dialogues, which are joined by a dialogue identifier and ordered by turn. Every row carries the public URL of the page on which that sentence is published, so any row can be traced back to the text a learner actually reads.

Provenance. The corpus is the material layer of a free, advertisement-free English learning site built by one parent for one nine-year-old child, and it is extracted programmatically from the published pages rather than maintained separately, so the dataset and the deployment cannot drift apart silently. The child has never attended a private English academy and has never used a commercial workbook. Korea has an unusually large private English-education market, which makes a fully documented record of a home-only path - the sentence inventory, the level mapping, the language coverage - an artefact others can inspect, reuse and disagree with, rather than an anecdote.

Language coverage. Coverage follows the audience of each collection. The US-curriculum, everyday-conversation, books-and-media, video-game and advanced-topic collections carry all ten languages, because they are written for learners anywhere. The Korean-school-curriculum collections carry English, Korean, Japanese and Chinese: these 1,488 sentences follow the national curriculum of one country and are written for learners in Korea and its immediate neighbours. The languages_filled column records this per row.

Use of AI tools. The English sentences, the level mapping, the section structure and the editorial decisions are the author’s own and are not machine output. The nine-language glosses were produced with the assistance of large language models and reviewed by the author before publication; the author is a native speaker of Korean and writes the Korean glosses directly. The website’s code and the extraction script that produced these files were written with the assistance of an AI coding assistant, reviewed and run by the author. No AI system is listed as an author or contributor; responsibility for all contents rests with the named author.

Files. sentences.csv and sentences.json (13,133 rows), vocabulary.csv and vocabulary.json (647 rows), and README.md, which carries the full column dictionary, the build method and the disclosure above. All files are UTF-8. Licensed CC BY 4.0.