The file I was working with is very large, even after removing some chapters. Word crashed my computer several times when I was working on some of the global edits (specifically clearing format and removing hyperlinks), so I resorted to selecting about 30 pages at once and performing those tasks only in those, before moving to the next block.
Having a reader that conveys accurate information about all the formatting in the epub is essential. I think that the biggest challenge is that there does not seem to be a way to see some of the format of the epub after converting it into other file types. The formatting changes to accommodate for the parameters of that software. For example, the heading levels that JAWS read in the .docx file was different from the html. I examined two html versions: one that I produced from the original epub using Pandoc, as well as unzipping the epub file (by changing the extension from epub to zip, and then opening the files within the .zip file. I could not find any of the block quotes either in the html or docx versions.
The book I was working with had several level four headings that were not formatted as headings; they appeared in bold text. But because I had first cleared all the format in Word, I could not find bold text. Also, many of the – that precede the name of an author after a block quote did not transfer to the .docx file at the time of conversion either; I could find some quotes by searching “–“ but not all of them.
I found that Word can be rather frustrating when it decides for the user. For example, I tackled the task of removing the decorative images globally. I found them, selected and replace them with three asterisks. I did not realize at that time that in many instances Word simply removed the asterisks, only when I was working in another task I saw they were missing. I assume that it saw them as misspellings, because the line disappeared entirely. I had not set Word to not autocorrect, this issue may be solved doing so.
Also, although for the most part the smaller images are decorative, there were a few instances of smallish images that I thought were decorative when they were not. This is another example of information that is nod conveyed by JAWS.
Some of the captions for images were read by JAWS as regular text, I did not know they were captions, their content did not make it apparent that they were not part of the regular text, but captions.
The images that are not decorative that go into the edited file need to be the original images from the epub. I could find the images within the epub by unzipping it and looking for them into the folder, but without sighted assistance I could not know what each of the .jpeg files was.
I learned a lot about Word and feel much more confident using it, especially with the help of JAWS Text Analyzer, as well as JAWS' skim reading capability (thanks Ka). I used many keyboard shortcuts and that made the process much simpler. I also learned to apply several styles in the style pane; some were hidden at first and so I could not find them when I pressed Control+Shift+S but when I checked the option to show all styles, and not only those in the document, I found them again and was able to apply them.
I also learned about wildcards, specifically "^$^p" to find headings without having to rely on the format. I plan to experiment more with wildcards, I am sure they will be useful in the future.
JAWS quick keys capability was very useful (Insert+Z) to find images with g, tables with t. After formatting the headings I could move through the Word document by headings (h) very easily.