Prepare a pronunciation question that a speaker or reviewer can evaluate in context.
A pronunciation review is most useful when the original input, generated representation, and speaker evidence remain connected. A screenshot of output alone may not explain the disputed sound or the meaning of the word.
| Part | What to retain |
|---|---|
| Input | Exact original spelling |
| Meaning | The intended sense or expression |
| Output | Practical notation and IPA candidate, identified separately |
| Question | The particular sound, syllable, or mapping you want clarified |
| Evidence | A contextual recording or speaker explanation where available |
| Proposed interpretation | What you think needs changing and why |
Say which part is uncertain. If you have no recording or independent example, a clearly stated question is still useful; do not invent evidence to complete the record.
A mismatch may concern the written input, a practical notation convention, an IPA mapping, or an exception in a particular word. Those are different questions. Preserve the original result so the discussion can identify which layer is involved.
A rule label in the experimental tool has a narrower scope than full validation of every possible word. Read it with its documented status and examples.
Identify what was spoken and its context. Retain the relevant point in a longer recording. Use material that is appropriate and agreed to be shared, following the same principles as Tabula’s recording guide.
Do not silently replace the original spelling with your interpretation of a sound. A transcription proposal should remain identifiable as a proposal until reviewed.
Record the explanation that resolved the question, or what evidence remains needed. The checked dashboard’s review-history and persistence features remain roadmap work, so a visible queue does not establish that a permanent public review record exists.
Read generation and review for the overall approach and questions and limits for the tool’s experimental boundaries.
