I am having some problems with the dictionary I am creating on my own from 7a
cmudict. (cmudict.0.7a)
For the compound words I just concatinate the corresponding phonetics from the
current dictionary.
This one fails (s3_decode_init exits)(-lts_mismatch yes or no still fails)
SPACE-ZOOM-OUT-SLOW S P EY1 S Z UW1 M AW1 T S L OW1
this one works
SPACE-ZOOM-OUT-SLOW S P EY S Z UW M AW T S L OW
These work
SPACE-ZOOM-IN-FAST S P EY1 S Z UW1 M IH0 N F AE1 S T
SPACE-ZOOM-IN-FAST(1) S P EY1 S Z UW1 M IH1 N F AE1 S T
SPACE-ZOOM-IN-MEDIUM S P EY1 S Z UW1 M IH0 N M IY1 D IY0 AH0 M
SPACE-ZOOM-IN-MEDIUM(1) S P EY1 S Z UW1 M IH1 N M IY1 D IY0 AH0 M
SPACE-ZOOM-IN-SLOW S P EY1 S Z UW1 M IH0 N S L OW1
SPACE-ZOOM-IN-SLOW(1) S P EY1 S Z UW1 M IH1 N S L OW1
SPACE-ZOOM-OUT-FAST S P EY1 S Z UW1 M AW1 T F AE1 S T
SPACE-ZOOM-OUT-MEDIUM S P EY1 S Z UW1 M AW1 T M IY1 D IY0 AH0 M
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am having some problems with the dictionary I am creating on my own from 7a
cmudict. (cmudict.0.7a)
For the compound words I just concatinate the corresponding phonetics from the
current dictionary.
This one fails (s3_decode_init exits)(-lts_mismatch yes or no still fails)
SPACE-ZOOM-OUT-SLOW S P EY1 S Z UW1 M AW1 T S L OW1
this one works
SPACE-ZOOM-OUT-SLOW S P EY S Z UW M AW T S L OW
These work
SPACE-ZOOM-IN-FAST S P EY1 S Z UW1 M IH0 N F AE1 S T
SPACE-ZOOM-IN-FAST(1) S P EY1 S Z UW1 M IH1 N F AE1 S T
SPACE-ZOOM-IN-MEDIUM S P EY1 S Z UW1 M IH0 N M IY1 D IY0 AH0 M
SPACE-ZOOM-IN-MEDIUM(1) S P EY1 S Z UW1 M IH1 N M IY1 D IY0 AH0 M
SPACE-ZOOM-IN-SLOW S P EY1 S Z UW1 M IH0 N S L OW1
SPACE-ZOOM-IN-SLOW(1) S P EY1 S Z UW1 M IH1 N S L OW1
SPACE-ZOOM-OUT-FAST S P EY1 S Z UW1 M AW1 T F AE1 S T
SPACE-ZOOM-OUT-MEDIUM S P EY1 S Z UW1 M AW1 T M IY1 D IY0 AH0 M
You also need ot remove numbers after phone names that mark stress.
so compound words will be less accurate?
I'm not sure where did you get this idea.
If compound words don't have stress marks wouldn't that make them less
accurate?
I'm trying to do a small command and control app and compound words seem to be
the only way to do it.
So i'm trying to get as much accuracy as possible.
Hence the stress marks on each individual word seems significant.
no
Definitely no, moreover compound words decrease accuracy.
You probably need to have different plan :)
If compound words are not the way to g, what other choices are there for a
list of phrases for command and control?
Finite state grammar is a simple way to describe small vocabulary language.
You can write finite state grammar in jsgf format.
If you are looking for more flexibility, you could build trigram language
model and implement semantic analyzer on top of it.