This is a very dirty and simple way to do some manipulation of endnotes in .docx files. This uses some of the work done on https://github.com/python-openxml/python-docx.git. However, it uses an old pull request from 2014 which was never merged, so that submodule commit has been added here.

Furthermore, it python-docx (at least in 2014) does not handle hyperlinks properly. This has been addressed in a very dirty way: A run in a .docx file is a stretch of text, and looks like:

<w:r>
    <w:t>hello</w:t>
</w:r>

while a hyperlink is the same (or more than one run), but wrapped in a <w:hyperlink> tag. The solution are the scripts clean_endnotes.sh and remove_links.py, which collectively remove all of the hyperlink tags.

First (and optionally), one can remove hyperlinks from an existing documents with

$ bash clean_endnotes.sh file.docx

which will produce a file called file-no-links.docx.

Finally, one should run

$ python3 endnotes.py file-no-links.docx

The default behavior of endnotes.py is to take the (at most) first 10 words leading up to the reference in the main text, and to add those to the start of the endnote reference.

endnotes.py has a few options. These are:

-n num - modify the number of words to keep leading up to the reference

-o file - choose an output file. The default is file-endnotes.docx if the input is file.docx.

--chapter num - Specialized. This will select only the references that are in Chapter num. Note that this is very article-specific, and probably won't work for yours.

--sort - Sorts the references. Default behavior is that the references occur in the resulting document in the order that they occurred in the original document; this will sort them alphabetically instead.