LLM human data · Code-Switching ASR
Human Data for LLM Projects for Code-Switching ASR
Human-generated text and speech for LLM training and evaluation in Indian languages: prompts, preference rankings, instruction-response pairs, red-teaming and cultural-fit review. Applied to code-switching asr, the specification is driven by one thing: switch-point accuracy.
- Service
- LLM human data
- Use case
- Code-Switching ASR
- Primary metric
- Switch-point accuracy
Required data profile
- Genuinely code-mixed spontaneous speech
- Per-token language ID labels
- A fixed rule for script of English tokens
Technical specification
| Parameter | Standard |
|---|---|
| Task types | Prompt writing, response ranking, instruction-response pairs, adversarial testing |
| Languages | Any language in the network, including code-mixed Hinglish |
| Contributors | Screened by domain, education band and language proficiency |
| Agreement | Overlapping assignments with adjudication |
| Provenance | Per-item contributor and time records |
Process
- Task specification and rubric design
- Contributor screening against the rubric
- Calibration round with feedback
- Production with overlap and gold items
- Adjudication and delivery
Metrics this feeds
- Switch-point accuracy
- Mixed-utterance WER
- Language ID token accuracy
Failure modes to design out
- Concatenating monolingual data and calling it code-mixed
- Leaving script conventions to individual annotators
Every item is traceable to a screened contributor, which matters when a model vendor audits your data provenance.
Deliverables
- Task data in your schema
- Rubric and calibration results
- Contributor metadata (anonymised)
- Agreement statistics
Frequently asked
Is llm human data the right service for code-switching asr?
It covers genuinely code-mixed spontaneous speech. Most code-switching asr programmes combine it with at least one other service; we will say so in the scope rather than selling one line item.
What languages are available?
All 14 languages in the network plus Indian English accent bands.
How is the evaluation set handled?
Collected first, from speakers disjoint from the training cohort, so improvement is measurable.
Scope llm human data for code-switching asr
Send the metric you need to move.