Add pocketsphinx_to_textgrid utility for Praat TextGrid conversion

[lenzo-ka]

Description

Add pocketsphinx_to_textgrid utility to convert PocketSphinx JSON alignment output to Praat TextGrid format.

Closes #455

Features

• Converts alignment JSON from pocketsphinx align to TextGrid format
• Supports all alignment levels: words, phones, and states (depending on -phone_align / -state_align options used)
• Outputs to stdout by default, or to a file with -o
• Optional Praat integration: -w audio.wav --praat opens the result in Praat

Usage

# Basic conversion to stdout
pocketsphinx align audio.wav "hello world" | pocketsphinx_to_textgrid

# Write to file
pocketsphinx -phone_align yes align audio.wav "hello" | pocketsphinx_to_textgrid -o output.TextGrid

# Open in Praat with audio
pocketsphinx -state_align yes align audio.wav "hello" | pocketsphinx_to_textgrid -o out.TextGrid -w audio.wav --praat

Verification steps

1. Build and install the package
2. Run alignment and pipe to the utility:

   pocketsphinx -phone_align yes align test/data/goforward.raw "go forward ten meters" | pocketsphinx_to_textgrid

3. Verify TextGrid output has words and phones tiers
4. Optionally open in Praat to verify alignment visually

[dhdaines]

Very cool and useful (I am almost certain that I wrote this myself at least 2-3 times in the past, but no longer know what happened to those...)

[dhdaines]

Ah in fact they probably are in https://github.com/ReadAlongs/Studio somewhere. There's also code there to output ELAN EAF files, which may be useful here too.

[lenzo-ka]

Might be worth consolidating closer to the aligner?

[dhdaines]

Might be worth consolidating closer to the aligner?

Hi! Really sorry about my super late replies here. Yes, I think this might be more useful as simply an option to pocketsphinx align.

TextGrid (unlike EAF) is a pretty simple format and it looks like this could be rewritten easily in C.

[dhdaines]

As mentioned in the comment! I think this should be straightforward to write in C and simply add as an option to pocketsphinx align. Particularly since we don't need to parse any JSON in that case.