German Special Characters

[JoHoenk]

Hi Guys,

first of all: awesome work here!

Currently I´m facing following issue with the special character 'ß' in sentence template:

brightness_numbers:
    values:
      - in: fünf
        out: 5
      - in: zehn
        out: 10
      - in: fünfzehn
        out: 15
      - in: zwanzig
        out: 20
      - in: fünfundzwanzig
        out: 25
      - in: dreißig
        out: 30
      - in: fünfunddreißig
        out: 35

When I run Vosk I get the following debug log entry, and the word recognition for 30 is also not working:
vosk | WARNING (VoskAPI:UpdateGrammarFst():recognizer.cc:308) Ignoring word missing in vocabulary: 'dreissig'

However if I look in the dictionary on Huggingface it looks like the work dreißig is included in the model.
So it seems like something is replacing the ß. I already tried out different quotes but I always get this warning.

Does anybody have a hint for me?

[Scrath1]

I experienced the same issue. The way I fixed it for me is the same way that the lumos example is described in the README file.
I took a look at the vocabulary for the german model (see here) and used other words or syllables to sound out the word.

For your example:

      - in: drei zig
        out: 30
      - in: fünf und drei zig
        out: 35

I do however still experience the same problem with the word "schließe" for which I haven't found a proper substitute yet.

[synesthesiam]

I think I see what's happening. I'm using Python's casefold by default, which turns "dreißig" into "dreissig". I think I should be using lower instead here. What do you think?

[JoHoenk]

That sounds right. This would probably solve the issue!