> With the UTF-8 attribute i have the correct visualization of this
> character but not yet for the chinese and arabic.

IMHO, the character encoding is correct. But Your text editor seems to
be unable to represent the non-latin characters.

I think too that this is the correct encoding, but in fact isn't,
anyway i don't open my result file in a text editor,
i use it for importing the terminology records in an old version of Multiterm by Trados software. 

Maybe an output encoded in ISO-8859-6 would give some results with arabic ?

I have tried to set this encoding but nothing is changed i can't see the arabic characters.

Find some test documents there ;-)
http://cvs.savannah.nongnu.org/viewcvs/aramorph/src/java/gpl/pierrick/brihaye/aramorph/test/testdocs/?root=aramorph

Tank's again,

Giuseppe.