[DECtalk] User dictionary
Kyle
kyle at free2.ml
Fri Feb 24 15:19:15 EST 2023
More information about my original issue:
I tried the udic_us and dic_us tools in my build to build the .tab file
into a user dictionary. The dic_us tool indicates that my dictionary is
being built with 0 entries and segfaults leaving a temporary file behind
and an empty dic_us.dic file. The udic_us tool completes with no errors,
produces by default a .dtu file with the same base filename as the .tab
file, and I can run strings on the file and see the word whatchamacallit
that I put in the .tab file with its correct pronunciation. The data
file does include some non-printable characters, whitespace and other
things to the right of whatchamacallit. I have tried converting the .tab
file between DOS and Unix formats before compiling. In all cases, I
loaded the compiled user dictionary into the say utility like so:
say -d dictionary.dtu
At this point I enter the word whatchamacallit and hear the original
pronunciation rather than my corrected version that it should see in my
user dictionary. So it seems that compiled dictionary files are ignored,
or else .tab files are not being compiled properly. Maybe this is a
question for anyone contributing to the code, since as I understand it,
singing was broken and now is fixed, so this may just be a bug, either
in the dictionary build tool or in the code that loads the user
dictionary into the say utility.
Looking further into this, I am seeing that say -h indicates that -d is
supposed to specify the audio device, 0 or 1, but the documentation I
find at
https://dectalk.github.io/dectalk/idh_nt_apps_dictionary_options.htm
tells me that -d is supposed to load the specified user dictionary file.
So it seems we have a conflict somewhere. It does in fact appear to be a
bug. Since I have no Github account, I am not able to report it,
although I am able to watch it periodically if someone reports it.
~Kyle
More information about the Dectalk
mailing list