fstmakecontextfst produces C.fst with duplicated arcs on disambig symbols

[taomanwai]

For this toy example, mono-phone of Chinese, where one phone exactly equal to one Chinese character

Why and is it a bug ?

[danpovey]

I don't know, it depends what args you called the program with.

…

On Sun, Jan 5, 2020 at 1:33 PM Tommy Tao ***@***.***> wrote: For this toy example <https://tommytaobucket.s3.amazonaws.com/C.pdf>, mono-phone of Chinese, where one phone exactly equal to one Chinese character Why and is it a bug ? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#3810?email_source=notifications&email_token=AAZFLO4FAADJ4INC5WABDTLQ4FWLJA5CNFSM4KCZSIK2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IEBXYRQ>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAZFLO4OTVFBWGSOAD7IUEDQ4FWLJANCNFSM4KCZSIKQ> .

[taomanwai]

Sorry for missing command line code

fstmakecontextfst --context-size=1 --central-position=0 --read-disambig-syms=lang/phones/disambig.int --write-disambig-syms=disambig_ilabels_1_0.int phones.txt 11 ilabels_1_0 > C.fst

PS:
phones.txt:
0
SIL 1
SPN 2
地 3
鐵 4
去 5
#0 6
#1 7
#nonterm:entity 8
#nonterm_begin 9
#nonterm_end 10

lang/phones/disambig.int:
6
7
8
9
10

[danpovey]

The following lines:

    for (size_t i = 0; i < disambig_in.size(); i++) {
      int32 sym = disambig_in[i];
      loop_fst.AddArc(0, StdArc(sym, sym, TropicalWeight::One(), 0));
    }

need to be removed from fstmakcontextfst.cc since those symbols are already added from phones.txt. Do you have time to make a PR:

[taomanwai]

Sure and wait, forking ...

[taomanwai]

PR created