| --- |
| configs: |
| - config_name: lexicon |
| data_files: |
| - split: terms |
| path: lexicon/terms.jsonl |
| - split: actions |
| path: lexicon/actions.jsonl |
| - config_name: mt |
| data_files: |
| - split: train |
| path: mt/train.jsonl |
| - split: validation |
| path: mt/validation.jsonl |
| - split: test |
| path: mt/test.jsonl |
| - config_name: asr |
| data_files: |
| - split: train |
| path: asr/train.jsonl |
| - split: validation |
| path: asr/validation.jsonl |
| - split: test |
| path: asr/test.jsonl |
| --- |
| |
| # OneVoice |
|
|
| Vietnamese-English garment-factory lexicon, MT, and ASR data exports. |
|
|
| ## Configs |
|
|
| - `lexicon`: seed terminology and action phrases. |
| - `mt`: train/validation/test JSONL from reviewed draft template sentences. |
| - `asr`: generated Vietnamese TTS WAV manifests, or empty placeholders when no generated manifest is configured. |
|
|
| ## Field glossary |
|
|
| - `text_vi`: Vietnamese factory-floor sentence. |
| - `text_en`: English translation of `text_vi`. |
| - `term_vi` / `term_en`: Vietnamese/English seed term in the lexicon files. |
| - `category`: seed term group, such as `equipment`, `material`, `process`, or `command`. |
| - `source_term`: source term or action phrase used to generate an MT sentence. |
| - `source_category`: category of `source_term` for action-paired MT rows; blank for plain lexicon rows. |
| - `paired_term`: object/error/QA term paired with an action phrase; blank when the sentence comes from a single plain lexicon term. |
|
|
