Query regarding Taggers and Verbalizers in RuleBasedITN

[DurgaiVS]

Hello everyone,

I hope this message finds you well. I have a question regarding Rule Based ITN flow.

On Rule based Inverse Text Normalisation. the transformation of text along with the tag was done by the Taggers FST itself, and then it reaches Verbalizers FST. This Verbalizers FST removes the tags in the transformed text and return the rest of the transformed text.

My question is, since the output of inverse_normalization is just the plain transformed text itself and no informations of the tag reaches out of the module,

• why can't the Taggers FST just tag the transformation region of a particular case, like, "tokens { measure { cardinal { integer: "fourty" } units: "percentage" } }" and hand over that particular region to that respective Verbalizers FST to do the transformation?
• Or, the other way around, like, why can't the Taggers FST alone is enough without their tagging feature but just the transformations, like, "fourty" -> "40" ?
• Why was this design chosen, any particular reasons?

Thank you for your time and support. I really appreciate the work and am looking forward to any suggestions or guidance you can provide.

Best regards,