Fix mkdir bug in speech_train_punkt_tokenizer.py

gooofy / zamia-speech

Open tools and data for cloudless automatic speech recognition

GNU Lesser General Public License v3.0

443 stars 84 forks source link

Fix mkdir bug in speech_train_punkt_tokenizer.py #86

Closed mrezai closed 4 years ago

mrezai commented 4 years ago

Fixed by define PUNKT_PICKLEFN as "Path" in parole.py.

gooofy commented 4 years ago

how would one reproduce the mkdir bug you're referring to?

mrezai commented 4 years ago

Run ./speech_train_punkt_tokenizer.py From Readme: sentences extraction: train punkt tokenizer using ./speech_train_punkt_tokenizer.py The problem is in this line: https://github.com/gooofy/zamia-speech/blob/0891e90afdbc09e49f1e440b29daf3233cf131b5/speech_train_punkt_tokenizer.py#L71 This line will be run only if "PUNKT_PICKLEFN" be defined as "Path" not string

gooofy commented 4 years ago

ah, I see - thanks for the explanation. Since I would like to keep dependencies low (esp. to packages that are not in CentOS 7) I decided to use misc.mkdirs instead. Turns out I could also get rid of the plac dependency as well.

Issue should be fixed now.