Question about mixed languages recognition (user-words are present and called by --user_words_suffix=user-words, but seems to be unworkable)

[Alexey Kostylev]

First of all - sorry 4 possible mistakes, English is not my native language indeed...

Used pytesseract with version

tesseract 4.0.0-alpha.20170804

 leptonica-1.74.4

  libgif 4.1.6(?) : libjpeg 8d (libjpeg-turbo 1.5.0) : libpng 1.6.20 : libtiff 4.0.6 : zlib 1.2.8 : libwebp 0.4.3 : libopenjp2 2.1.0

parameters seems 2 be OK

config:

lang = eng+rus
tess_config = --tessdata-dir "D:\Program Files (x86)\Tesseract-OCR\tessdata" --psm 4 --user_words_suffix=user-words
tesseract_cmd = D:\Program Files (x86)\Tesseract-OCR\tesseract.exe

command:

text = pytesseract.image_to_string(Image.open("temp/BW.PNG"),lang=self._lang, config=self._tess_config )

User words files are exactly present,

D:\Program Files (x86)\Tesseract-OCR\tessdata>dir *.user-words

...

Содержимое папки D:\Program Files (x86)\Tesseract-OCR\tessdata

10.02.2018  09:40               169 eng.user-words

09.02.2018  22:21                 9 rus.user-words

but looks like, that the files are not loaded, or, at least, used not effectively. Here are the images (original and BW, cleaned by me)

Part of eng.user-words (was trying the words, starting from uppercase too)

societas

eruditoium

civitas

dei

icu

ino

rodentia

earth

expeditionary

pyrrha

santos

dumont

The result from OCR is below, lines with names from user-words file are made bold by me. 

Rock Research Ring

Текущее влияние: 1%

В ближайшем будущем ожидается изоляция.

В ближайшем будущем у них ожидаются гражданские

беспорядки.

Планируется экспансия из ВодепНа Petram.

В состоянии войны с Еагіб Нее в системе

Phoenix.

Их ждет экономический бум.

Colonia Council

Текущее влияние: 4%

В состоянии войны с босіетаз Егидиогит де Суйаз Реі в

cucteme Dubbuennel.

Any advice? Seems 2 me, that I need to force user dictionary over native, or add values for mixing languages, but how? 

P.S. I am complete noob with tesseract, but very impressed indeed. Nearly all regular text from PC screenshots is parsed OK