Instead of using WordExtractor, you can read with Range:
...
HWPFDocument doc = new HWPFDocument(fis);
Range r = doc.getRange();
...
Range is the central class of that model. When you get range, you can play more with the features of the texts and, for instance, iterate through all CharacterRuns, and check if it is Italic (.isItalic()) or change to Italic: (.setItalic(true)).
for(int i = 0; i<r.numCharacterRuns(); i++)
{
CharacterRun cr = r.getCharacterRun(i);
cr.setItalic(true);
...
}
...
File fon = new File(yourFilePathOut);
FileOutputStream fos = new FileOutputStream(fon);
doc.write(fos);
...
It works if you are stick to use HWPF. Between, to frame into and work with the concept of Paragraph is more convenient.