Downloads · 30 days
25
25% of all-time downloads
korean-spacing-ONNX
noticemkjung/korean-spacing-ONNX
korean-spacing-ONNX is a token classification model from noticemkjung. Use it when you need labels on individual words, such as names. It is set up for transformers.js. The card lists the license as apache-2.0.
ONNX int8 export of fiveflow/roberta-base-spacing, for use in the browser with transformers.js.
All-time downloads
99
Public
Repo size
111 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx111 MB · 99%
At a glance
- Task
- Token Classification
- Library
- transformers.js
- License
- apache-2.0
- Model type
- roberta
- Access
- Public
- Created
- Aug 22, 2026
- Updated
- Aug 22, 2026
- SHA
- 10f8ed13
Base models
Model card
From the Hugging Face model README
korean-spacing-ONNX
ONNX int8 export of fiveflow/roberta-base-spacing, for use in the browser with transformers.js.
Used by an offline voice translator to repair Korean word spacing in
speech-recognition output: some recognizers get the words right and split them
in the wrong places ("uc6b0 ub9acub098 ub77cub294" for "uc6b0ub9acub098ub77cub294"). Feed the text with
spaces removed, one character per token, and insert a space after every E tag.
- 440MB fp32 -> 110MB int8
- Quantized graph verified against the original: identical spacing on all test sentences.
- Task
- Token Classification
- Library
- transformers.js
- Type
- roberta
- License
- apache-2.0
- Languages
- ko
- Created
- Aug 22, 2026
- Updated
- Aug 22, 2026