Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,50 @@
# Fixing Misspelled Words in Deepgram Nova-3 en-GB Transcriptions

## Addressing Misspelled Words in Deepgram Transcriptions

Misspelled words in audio transcriptions can be a frustrating issue, impacting the accuracy and usability of the generated text. When using the Deepgram Nova-3 en-GB streaming through the WebSocket API, users may occasionally encounter misspelled words in their transcriptions, such as 'collabouration', 'manageing', and 'maxeimizing'.

By understanding and addressing potential causes, users can improve transcription quality and reduce the occurrence of such issues.

### Investigating Configuration Parameters

One potential cause of misspelled words in transcriptions is incorrect configuration of query parameters when calling the Deepgram API. A common oversight is related to the `redact` parameter, which controls redaction during transcription.

#### Example Issue

An example log error that may correlate with transcription problems is:

```
Error: Query parameter redact was set to ['false']. Input should be a valid string.
```

To enable proper redaction, the query string parameter should be correctly configured in your API call:

- **Redaction Parameter**: Ensure you set a valid parameter value when making requests to Deepgram's `/listen` endpoint:

```
redact=OPTION
```

For more information on valid options, see the [Deepgram Redaction Documentation](https://developers.deepgram.com/docs/redaction).

### Suggestions for Troubleshooting

1. **Review Query Parameters**: Double-check query parameters you're using when calling the transcription API to ensure all values are valid according to the [Deepgram API documentation](https://developers.deepgram.com/docs/getting-started-with-live-streaming-audio).

2. **Redaction Settings**: If you use redaction, ensure these parameters are explicitly defined and align with best practices. Misconfigurations can lead to unexpected transcription results.

3. **Testing**: Isolate and test different components of the system independently to ascertain if the issue persists across configurations or is specific to particular settings.

4. **Consult Support & Community**: If issues persist or the system behavior seems inconsistent, reach out to your Deepgram support representative (if you have one) or visit our community for assistance: [Deepgram Community on Discord](https://discord.gg/deepgram).

### Conclusion

By ensuring you adhere to correct configuration values and continually testing with different settings, you can improve the quality of transcriptions from Deepgram's Nova-3 en-GB model.

For a robust integration, validating each parameter and iterating on settings can help resolve many common issues related to transcription quality.

### References

- [Deepgram Redaction Documentation](https://developers.deepgram.com/docs/redaction)
- [Live Streaming Audio Documentation](https://developers.deepgram.com/docs/getting-started-with-live-streaming-audio)