Source files directory for the trainer
--------------------------------------
Create a directory for each language to model, using the identifier for the
language as the name for the directory.
You are encouraged to use ISO 639-1 language codes (es,en,de,fr, etc.) but you
can use the names you want (spanish, english, german, french, ...) The trainer
will use blindly the directory name as the identifier for the language.
So, if you use "alemán" as the name of the directory with the german train data,
the library will identify texts like those as "alemán", not "german", nor "de".
Into every directory, copy sample texts for the language. Encode all of them in
UTF-8 only, and use only plain text files with .txt extension (or .txt.gz if
you want to save space).
After running the trainer, the models for every language will be saved in the
"model" directory.
Good luck
|
Download
