combine_lang_model - Man Page

generate starter traineddata


combine_lang_model --input_unicharset filename --script_dir dirname --output_dir rootdir --lang lang [--lang_is_rtl] [pass_through_recoder] [--words file --puncs file --numbers file]


combine_lang_model(1) generates a starter traineddata file that can be used to train an LSTM-based neural network model. It takes as input a unicharset and an optional set of wordlists. It eliminates the need to run set_unicharset_properties(1), wordlist2dawg(1), some non-existent binary to generate the recoder (unicode compressor), and finally combine_tessdata(1).


--lang lang

The language to use. Tesseract uses 3-character ISO 639-2 language codes. (See LANGUAGES)

--script_dir PATH

Directory name for input script unicharsets. It should point to the location of langdata (github repo) directory. (type:string default:)

--input_unicharset FILE

Unicharset to complete and use in encoding. It can be a hand-created file with incomplete fields. Its basic and script properties will be set before it is used. (type:string default:)

--lang_is_rtl BOOL

True if language being processed is written right-to-left (eg Arabic/Hebrew). (type:bool default:false)

--pass_through_recoder BOOL

If true, the recoder is a simple pass-through of the unicharset. Otherwise, potentially a compression of it by encoding Hangul in Jamos, decomposing multi-unicode symbols into sequences of unicodes, and encoding Han using the data in the radical_table_data, which must be the content of the file: langdata/radical-stroke.txt. (type:bool default:false)

--version_str STRING

An arbitrary version label to add to traineddata file (type:string default:)

--words FILE

(Optional) File listing words to use for the system dictionary (type:string default:)

--numbers FILE

(Optional) File listing number patterns (type:string default:)

--puncs FILE

(Optional) File listing punctuation patterns. The words/puncs/numbers lists may be all empty. If any are non-empty then puncs must be non-empty. (type:string default:)

--output_dir PATH

Root directory for output files. Output files will be written to <output_dir>/<lang>/<lang>.* (type:string default:)


combine_lang_model(1) was first made available for tesseract4.00.00alpha.


Main web site: Information on training tesseract LSTM:

See Also



Copyright (C) 2012 Google, Inc. Licensed under the Apache License, Version 2.0


The Tesseract OCR engine was written by Ray Smith and his research groups at Hewlett Packard (1985-1995) and Google (2006-present).