Regular expression for arabic text in matlab

I used ocr in matlab to read arabic text from image.Now I want to write a regular expression that matches a word in arabic text but it does not work

[Walter Roberson]

Could you give us some specific examples to experiment with?

[N Rh]

For example I want to extract the text "فاتورة عدد" in the attached image.

[Stephen23]

"...but it does not work"

This tells us nothing about what you have tried so far, nor about what the difference is between working/not working.

What have you tried? Do you get an output? How are you checking this output? If no output, do you get any error message?

[N Rh]

the output is "????????" it means that the regular expression does not support the arabic

[Guillaume]

"it does not work"

As Stephen's said this is a useless statement if you don't even tell us what the "it" is. How can we know if you've made a mistake with the "it", or if you're using the "it" incorrectly, or if indeed the "it" does not support arabic.

So show us the "it", that is the exact code you're using and ideally an example input where "it" doesn't work.

[N Rh]

The "it" means for example: Pattern=['(فاتورة عدد)']; P = regexp(Lines,Pattern,'match'); P=[P{:}];

[Guillaume]

Seems to work for me (R2017b):

>> Pattern = '(فاتورة عدد)';
>> Lines = {Pattern(2:end-1); [Pattern(2:end-1), '2015/02 ']; Pattern(4:5)}
>> P = regexp(Lines,Pattern,'match');
>> P = [P{:}]

Lines =

    3×1 cell array

      {         فاتورة عدد'}
      {'فاتورة عدد2015/02 '}
      {                'تو'}

P =

    1×2 cell array

      {'فاتورة عدد'}    {'فاتورة عدد'}

[N Rh]

Thank you i will install R2017b and try it.

[N Rh]

With R2017a I have this result:

Lines = 3×1 cell array

    '?????? ???'
    '?????? ???2015/02 '
    '??'

P =

1×9 cell array

    '??'    '??'    '??'    '??'    '??'    '??'    '??'    '??'    '??'

[Guillaume]

I don't think it is an issue with your matlab version as it also works for me in R2016a, R2016b and R2017a. It's probably more related to your operating system.

I'm using Win 7 (Enterprise) and didn't have to do anything special to get the above to work.

[N Rh]

Can you tell me how you use ocr in matlab for recognition

[Walter Roberson]

One thing to note is that if your operating system is set to English, then MATLAB might not store .m files with UTF encoding, so when you save the .m file and close it and open it again, any arabic characters you had in the file might be gone. With newer versions there is apparently a way to force MATLAB to permit UTF-8 for .m files, but it involves editing an obscure configuration file.

[Walter Roberson]

Just to be sure we are all referring to the same thing:

It is not possible to use regexp() on an image, only on character vectors or cell array of character vectors or on string() arrays.

[N Rh]

Yes i use regexp() on cell array i don't know what is the problem!!!

[Walter Roberson]

Please attach a .mat containing the cell array and also containing the pattern you are trying to search for.

[N Rh]

this is the used code, you can execute it and the image in the attached file.

clear all;close all;clc;
!tesseract -l eng+ara fac.jpg output 
slCharacterEncoding('UTF-8')
        fid = fopen('output.txt');              
        b = fread(fid,'uint8')';
        fclose(fid);
        a=dec2bin(b);
        c=dec2hex(b);
        str = native2unicode(b,'UTF-8');
        disp(str);
        C = textscan(str,'%s');
        data=cellstr(C{1});
         for i=1:length(data)
            if strfind(char(data(i)), 'عدد')==1 
                 fprintf('Numero de la facture : %s\n',char(str(i+1)))
             end
         end

[Walter Roberson]

I had to hunt around for the arabic training files for tessaract; perhaps I did not find the right ones. And I got a whole bunch of messages about

Cube ERROR (ConvNetCharClassifier::RunNets): NeuralNet is NULL

The output.txt file contained only English for me.

[N Rh]

because you need files ara.cube.bigrams, ara.cube.fold, ara.cube.lm, ara.cube.nn, ara.cube.params, ara.cube.size, ara.cube.word-freq, ara.traineddata

[N Rh]

The attached file contains the cell array of words: فاتورة, عدد, 2015/02