Cantonese speech dataset
We source Cantonese speech data to order: speaker recruitment, recording, transcription and annotation, delivered with the consent documentation that makes the dataset usable rather than merely delivered.
What makes Cantonese hard to collect
Cantonese has six to nine tones, and many colloquial words have no standard character at all: annotators end up improvising characters or falling back on romanization. The lack of a unified written standard is the biggest trap in this kind of corpus.
This is the part that cannot be templated. Every language has a specific failure mode, and a project that does not plan for it produces data that looks complete and trains badly.
Language profile
| Language | Cantonese |
|---|---|
| Primary region | East Asia |
| Writing system | Han (Traditional) |
| Collection difficulty | Tonal language |
What we deliver for Cantonese
-
Collection to your specification
Scripted or unscripted, in the language, domain and conditions your task needs. Specified separately where the variants need different contributor pools.
-
Labels to your convention
Orthographic, phonetic or categorical, with the guideline written before production and shared with you for review.
-
Contributor metadata
The attributes that let you slice the set rather than take it whole. The ones your task turns on, not a fixed list.
-
Consent documentation
Signed contributor consent covering the intended use, plus collection methodology and the annotation guideline.
-
Pilot batch
A small batch first, which you can reject. Misalignments surface after a few hours rather than at delivery.
-
Delivery in your format
File format, structure, segmentation and metadata schema set to your pipeline's requirements.
Questions we get about Cantonese
How many distinct speakers can you provide for Cantonese?
It depends on the specification and the timeline, and we will give a real number rather than a target. Speaker recruitment is usually the step that sets the schedule for Cantonese, so the answer to this question is also the answer to how long the project takes.
Can you annotate to a custom guideline?
Yes, and for Cantonese we would recommend it. The default conventions for this language leave several decisions open, and those decisions are exactly where annotators diverge. Send us your guideline, or we will draft one and you can review it before production.
Do you offer a sample before we commit?
Yes. Select a free sample in the request form and describe what you need. A pilot batch is the cheapest way to establish whether the quality bar is reachable before committing to a full run.
What if Cantonese is not something you can source?
Then we will tell you rather than take the order. We would rather decline a project than deliver something that does not meet the specification.
Other languages in East Asia
Request Cantonese speech data
Tell us the language, the hours, and what the data needs to look like. You will get a real number and a real timeline, not a range. If we cannot source it well, we will tell you that instead.
- Pilot batch before the full run, so problems surface early.
- Consent documentation delivered with the data.
- No medical or clinical data. No recorded telephone calls.
- You send the spec Language, hours, recording conditions, and the deadline.
- We reply within two business days A real number and a real timeline, not a range.
- You approve a pilot batch A small run first, so problems surface before the full order.
- Full run, then delivery with documents Audio, transcripts, and the consent record for every speaker.