IIRC, that is just 1:1 mapping between utterance name and a number. The index stores the id instead of the utterance name.
You can generate it any way you want (but I prefer having the utterances indexed sequentially), it will be needed only when processing the keyword search results.
Y.
--
You received this message because you are subscribed to the Google Groups "kaldi-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kaldi-help+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Yes, that should be fine - the format is similar to what words.txt or phones.txt look like
Y.