Created attachment 26514 [details]
MS Word document on which the effect can be reproduced.

When MS Word document (please see the attachment) containing greek characters is passed to org.apache.poi.hdf.extractor.WordDocument. Method writeAllText returns incorrect-incomplete result. No exception is thrown to indicate the problem.

Steps to reproduce:

1. Use the MS Word document from attachment.
2. Create the input stream of the document and then use this snippet:

            WordDocument wd = new WordDocument(inputStream);
            StringWriter docTextWriter = new StringWriter();
            PrintWriter pw = new PrintWriter(docTextWriter);
            wd.writeAllText(pw);
            result = docTextWriter.toString();

3. Expected result is string containing "Process description document τεστ new"
4. Actual result is "Process description"
5. No sign of internal error indicated, no exception is thrown.

I would expect at least exception thrown as an indicator that something went wrong.