What steps will reproduce the problem?
1. train a very small model
2. call phonetisaurus-g2p with a word as input which uses only "unknown"
symbols:
phonetisaurus-g2p --model=foo.fst --nbest=3 --input=foobar
What is the expected output? What do you see instead?
I do this during work on my diploma thesis when experimenting with "very small"
models: when all symbols of the input are mapped to null, phonetisaurus seems
to enter an infinite loop instead of bailing out / complaining / producing
empty output
What version of the product are you using? On what operating system?
I use phonetisaurus 0.7.8 on SuSE linux 12.3 (3.7.10-1.1-desktop, amd64)
Please provide any additional information below.
adding a check for an empty vector entry fixes the issue for me (see attached
patch)
Original issue reported on code.google.com by merz.mat...@googlemail.com on 17 Jun 2013 at 12:46
Original issue reported on code.google.com by
merz.mat...@googlemail.com
on 17 Jun 2013 at 12:46Attachments: