Japanese Phonetic (Romaji) Error Dataset

Full Official Name: Japanese Phonetic (Romaji) Error Dataset
Submission date: March 13, 2026, 7:59 p.m.

A corpus of Japanese sentences with programmatically introduced phonetic Romaji errors (Missing, Addition, NearbyKey Typo, TwoKey Misorder), designed for evaluating Input Method Editor (IME) systems. Derived from the Snow Simplified Japanese Corpus.

Creator(s)
CUNY Graduate Center - Yuchun Zou
CUNY Hunter College - Tedd Lee
Meta Inc. - Xiaodi Fan
CUNY Queens College & Graduate Center - Jun Li
Distributor(s)
CUNY Graduate Center - Yuchun Zou
CUNY Hunter College - Tedd Lee
Meta Inc. - Xiaodi Fan
CUNY Queens College & Graduate Center - Jun Li
Right Holder(s)
CUNY Graduate Center - Yuchun Zou
CUNY Hunter College - Tedd Lee
Meta Inc. - Xiaodi Fan
CUNY Queens College & Graduate Center - Jun Li
Status : Accepted
ISLRN :
482-587-871-494-1
Version
1.0
Source
https://github.com/junli-cuny/llm-ime-benchmark/tree/main/data/japanese_error_dataset
Resource Type
Primary Text
Media Type
Text
Language(s)
Japanese
Access Medium
Downloadable