Add more languages support for disfluency#289
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
giorgosHadji
changed the title
giorgosHadji
changed the title
petemomo
left a comment
petemomo
left a comment
There was a problem hiding this comment.
Thanks for this — and for adding the language codes, which brings the list in line with the profanities convention on the page. One ask, one question (inline), and two optional nits (inline).
The API spec still says English-only. spec/realtime.yaml (line 1130) reads "[disfluency] is only present in English", and the API reference pages are generated from that spec — merged as-is, the docs would claim both 20-language and English-only support. Since the spec lives in this repo, could you update that description in this PR? If the spec change is tracked elsewhere (e.g. it mirrors an upstream source), let me know and we'll file a follow-up instead.
|
|
||
| ### Supported languages | ||
|
|
||
| Disfluency tagging and removal are available for the following languages. Each language has its own set of hesitation sounds; the full list above is for English. |
There was a problem hiding this comment.
Nit: "the full list above" is a positional reference. Something like "the expandable list at the start of this section covers English" survives restructuring better.
There was a problem hiding this comment.
Indeed! Nice! changed
I have changed the spec as well here to reflect the change. |
petemomo
left a comment
petemomo
left a comment
There was a problem hiding this comment.
All addressed — thanks for the fast turnaround, and for going beyond the ask: catching the extra reference in voice-agent-api.mdx and making the Smart formatting heading self-contained too.
Verified locally on the updated branch:
- No English-only disfluency claims remain anywhere in docs or spec.
- The new
#supported-languages-for-disfluenciesanchor is used consistently in all three places, and nothing links to the old#languagesanchor, so that rename is safe. - The page renders correctly with the new headings and updated in-page link.
One optional nit, take it or leave it: "Disfluency detection is supported for [supported languages]" in voice-agent-api.mdx doubles up — "available for the [supported languages]" reads smoother.
The Mandarin naming unification stays as the agreed separate task — we'll pick it up on our side. LGTM 👍
…-up) (#294) Follow-up agreed in the #289 review (with @giorgosHadji): the docs should use **"Mandarin"** everywhere, not "Chinese Mandarin", matching the canonical row on the languages page (`Mandarin | cmn`). Targeting `feature/disfluency` so it rides in with #289. ## Changes **`docs/speech-to-text/formatting.mdx`** — the three "Chinese Mandarin" occurrences: - "Available Chinese Mandarin locales:" → "Available Mandarin locales:" - Disfluency supported-languages list: "Chinese Mandarin (`cmn`)" → "Mandarin (`cmn`)", moved to its alphabetical position (M) - Smart formatting languages list: "Chinese Mandarin (Simplified and Traditional)" → "Mandarin (Simplified and Traditional)", also moved to M **`.CLAUDE/context/terminology.md`** — records "Mandarin" as the canonical form ("Chinese Mandarin"/"Mandarin Chinese" in the do-not-use column). This is assistant-context, not deployed site content. ## Checks - Repo-wide grep (`docs/`, `spec/`, `src/`) — no other "Chinese Mandarin" occurrences remain - cspell on the changed file: 0 issues - No headings changed, so no internal anchors break 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Pete Mo <175202887+cabbage-ice-cream@users.noreply.github.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2 participants
No description provided.