Specific TextFragments that are positioned a little further from the previous text are incorrectly recognized as word start by the text recognizer engine and result in additional white spaces in the exported text.