Nexiwave Tips
There are ways you can help our engine to get better results.
Try these "obvious" things - use better audio
- Use better microphone (such as using noise cancelling microphone)
- Use less compression when recording
Choose correct decoding profile - use the "decodingProfile" parameter
"Decoding Profile" is configuration that configured specifically to work with your audio. For example, if the audio was voicemail, try to choose "Voicemail" as the decoding profile when submitting the audio.
Separate Your Content in Nexiwave - use the "sourceKey" parameter
If you are a multi-tenant system (eg, your Nexiwave account is used to hold content of many of your own users), it is suggested that you attach the optional "sourceKey" field when sending media content to Nexiwave.
Benefits:
- Nexiwave builds separate speaker profile database for each unique sourceKey. If your account contains too many speakers, it will be very difficult for our engine to separate speakers (think how hard it can be for a human to separate hundreds of speakers)
- Each search query sending to nexiwave can also have a "sourceKey" field. Queries with different sourceKey will not be applied to the indexing process of recordings with different sourceKey. Hence reducing the interference between your different users.
Supply "Contextual Data"
Believe it or not, we actually programmed our speech engine to work like human: it works much better if it knows some context. Imagine the trouble an art major student would have in an advanced mathematics class.
Contextual Data is the way for our engine to know the context. It helps tremendously on the accuracy.
When the media is being submitted to nexiwave, "Contextual Data" can be supplied using these optional fields:
- the "displayTitle" field, which can map to the subject of the talk.
- the "notes" field, which can be any comments, notes from the user
- the "keywords" field, which can be raw words/sentences extract from:
- PowerPoint Presentation for the talk
- Word document that drafted the play
- Any text that might are related to the speech content
Talk to us
There are often many ways to satisfy your users' needs. This is particularly true for complex speech-based system. We can advise you if you have utilized the best of our technology. Drop us a note and we sure can help out. It can make a big difference.
