If you notice the transcript imperfections or the speakers are not correctly recognized, make sure you troubleshoot using the below instructions.
- To improve the transcript and diarization quality, ensure the recording is clear, without background noises, and the language is understood. If Sembly cannot distinguish the speech from the surrounding noises, it may result in inaccurate transcripts;
- Speak clearly and naturally, at your regular conversational tone and pace;
- Speak for at least one minute without pauses or voice overlapping during the call to be diarized as an individual speaker;
- Avoid speakers interrupting one another. Sembly may not be able to identify overlapping voices correctly;
- Try using an external sound input device or a headset designed for online calls. Check out the Philips Smart Meeting device;
- Stay at close range of the microphone and avoid moving or touching the device to minimize any unnecessary background noise;
- Train Sembly to identify you by voice and automatically tag your name in the transcript by enrolling your personal Voice ID.
If you feel like all the requirements for the perfect transcript are met, but you notice the issues, you can apply for the revision of your meeting. If you do, fill out the consent form allowing us to access your meeting data.
Transcript (or speech to text) - a converted speech into text.
Speaker recognition - identifying individuals by their voices in audio.
Diarization - partitioning the transcript into segments according to the speaker's identity.
Utterance - a block of speech in the transcription. A long speech of a single speaker can be broken down into several utterances.