The output generated by this converter module depends a great deal on the input present in the word document. If the markup tags aren't mapped properly, headings are just marked with a big, bold font, but not with the appropriate tags for headings in HTML (<H1>, <H2>, ...), making it difficult if any post-processing of the output is necessary. It can be said in general that the conversion is rather poor, because mostly only paragraph boundaries are kept during the conversion. Changes in the font size, sub- and superscript are dropped, which is especially unpleasant for scientific documents. Images imbedded in the document are not converted, neither are tables and formulas.
No document segmentation is performed by the Word-Internet-Assistant, this is left to the converter doing this by hand. No need to mention what this does to any existing links within the document.
Some of be bugs described above are fixed in the Assistant V7.0, but by far not all.