Voice Assistant Companies · Code-Switching ASR
Code-Switching ASR Data for Voice Assistant Companies
Device, OS and appliance makers shipping assistants into Indian homes and vehicles, where wake-word reliability and far-field accuracy decide the review scores. Recognising speech that switches between an Indian language and English several times per sentence.
- Buyer
- Voice Assistant Companies
- Use case
- Code-Switching ASR
- Metric
- Switch-point accuracy
Where the two meet
Wake-word false accepts and rejects spike on Indian phonetics That is a code-switching asr problem, and it is solved by data shaped like this:
- Genuinely code-mixed spontaneous speech
- Per-token language ID labels
- A fixed rule for script of English tokens
Your evaluation criteria
- Can recordings be captured at the distances and conditions the device sees?
- Are entity-heavy prompt sets available (names, addresses, PIN codes, amounts)?
- Can negative wake-word data be collected alongside positives?
Metrics
- Switch-point accuracy
- Mixed-utterance WER
- Language ID token accuracy
Pitfalls
- Concatenating monolingual data and calling it code-mixed
- Leaving script conventions to individual annotators
Contract points
- Device-specific recording conditions
- Exclusive use of the collected wake-word data
- Staged delivery per firmware milestone
Frequently asked
What does a first engagement look like?
Usually a scoped pilot: one language, an evaluation set plus a first training batch, delivered in three to five weeks, followed by the full programme.
Can you match our existing vendor's schema?
Yes. Working to your schema avoids a conversion pass and keeps deliveries comparable across vendors.
How is provenance documented?
Per-item contributor records and consent mapped to IDs in the manifest.
Send your requirement
Language, volume, metric, deadline.