#23 Ambiguity -- please provide separator phonemes

[macias]

Currently phonemes are written as one string, for example:

espeak -x -q -z thai

results in a string

t'aI

However this is ambigous -- because espeak uses such phonemes (mnemonics) as "a", "I" and "aI".

Please add separator (spaces?) so the output would be:

t' a I

or

t' aI

This would eliminate ambiguity.

[Jonathan Duddington]

If you do:

espeak --ipa=1

it uses a tie character to join multiple characters within a phoneme name.

espeak --ipa=3

separates each phoneme with an underscore character.

Does this meet your need?

[macias]

Yes, thank you for your answer (I assume it works as you describe, because I have difficulties compiling the latest version right now).

[macias]

Ok, finally I got working espeak (latest released).

So my wish still stands -- ipa=3 indeed adds underscores, but it changes the phonemes as well.

My point is, there should be a way/option/default to add separators without changing anything else.

[Jonathan Duddington]

I can do this, but I'm not sure what is the best way to control it from the command-line interface.

1) Add an optional value to the -x and -X optiond, similar to the value after the --ipa option. The problem is that commands such as the convenient short-form:

espeak -Xvde

don't work, because the -v must now be a separate argument since the -X may have a parameter.

2) Use the --ipa=3 option to be used together with the -x and -X options, where

espeak -x --ipa=3

means produce ascii phoneme output with underscore separators. The --ipa option is used only to specify the separator, but the -x means that the output is ascii phonemes, not ipa phonemes. This is confusing.

3) Use a new option.

Sugestions?

[macias]

3.a) deprecate --ipa=3 smoothly
introduce new option (--sep?) with default value underscore

So ipa=3 would for now silently add this option, but user could override it with different separator.

"x" would work as always, unless user would provide either just "--sep" or with value for example "--sep=/".

In general move in the direction that producing phonemes in format X is independent concept of separating them.

[Jonathan Duddington]

Done in eSpeak 1.48.11, now at:
http://espeak.sf.net/test/latest.html

I added command-line options --sep (default is space) and --tie (default is U+361).

Details are in the updated:
http://espeak.sf.net/commands.html

[macias]

Thank you very much!