FastAlign/help.praat at main · jmlahoz/FastAlign · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
# FastAlign
# Author: José María Lahoz-Bengoechea
# License: GPL-3.0-or-later

clearinfo
printline Contact, questions, suggestions: José María Lahoz-Bengoechea'newline$''tab$'jmlahoz@ucm.es
printline
printline FastAlign creates 'phones', 'syll', and 'words' tiers from a Sound and a TextGrid
printline with an existing 'ortho' tier.
printline The 'ortho' tier can be automatically created with Whisper as a previous step,
printline as shown here: https://www.youtu.be/03EIj5UkJ5c
printline Then, use FastAlign as shown in this tutorial: https://youtu.be/O4CEjc63-qk
printline
printline -------------------------------------------------------------------------
printline
if windows = 1
printline Segmentation is based on HTK by default.
printline This draws from Hidden Markov Models (HMM).
printline FastAlign includes such a model especially trained with data in Spanish.
printline HTK alignment is more accurate than Praat native alignment functions.
printline However, sometimes it fails to produce a viable segmentation.
printline Only in those cases, native alignment will be invoked as a last resource.
printline Both algorithms are available to allow comparison and research.
elsif macintosh = 1 or unix = 1
printline Segmentation is based on Praat native alignment functions.
endif
printline
printline -------------------------------------------------------------------------
printline
if windows = 1
printline Run one-click segmentation with the function FastAlign.
printline
printline A step-by-step segmentation is also possible:
printline First, phonetize the ortho tier to create a temporary phono tier
printline (which consists of a phonetic transcription in SAMPA).
printline Then apply HTK alignment.
printline
printline As an alternative, just apply native alignment (no prior phonetization is needed in that case).
printline If you have an already segmented TextGrid and want to re-try segmentation of just one ortho interval,
printline you may choose to do so by running "Align sound (native)".
printline For just that one interval, information will be overwritten regardless of the overwrite checkbox.
elsif macintosh = 1 or unix = 1
printline If you have an already segmented TextGrid and want to re-try segmentation of just one ortho interval,
printline you may choose to do so by running "Align sound (native)".
printline For just that one interval, information will be overwritten regardless of the overwrite checkbox.
endif
printline
printline -------------------------------------------------------------------------
printline
printline In any case, alignment should be further hand-corrected.
printline Tip: hold SHIFT down while you click and drag any boundary of a given tier,
printline and the associated boundaries in the tiers below will move accordingly.
printline That way, you do not have to correct all tiers one by one,
printline and you can make sure that boundaries for different tiers will coincide as appropriate.