README-ODT

triggered by the recurring struggle to submit a paper to a conference that only accepts word, ...

I have been going through the respective PDF2Word tools, (e.g. the free http://www.pdftoword.com or the paid Adobe offering), and  the upshot is that they only allow you to rescue the text, an all other things have to be re-done manually.

Surprisingly, from a workflow perspective this is not much different from massaging the LaTeX sources in emacs and then loading them in a word file and then adding all the formatting by hand.

All of this is especially tedious for bibliographies/citations, footnotes, and figures (BCFF; after all, this is where we LaTeX users are especially spoilt).

The underlying problem is that all of these conversion pathways fail to preserve the semantics of the interesting parts (the BCFF) in the first conversion step.

But fear not, we have a tool that can help: LaTeX (you have guessed it), knows about the semantics. I tested this hypothesis and made a first LaTeX to OpenOffice converter based on LaTeXML. You can find it at https://svn.mathweb.org/repos/LaTeXML/branches/arXMLiv/tools/LaTeX2ODT (I did not quite know where to put it, please suggest something).

The converter runs LaTeXML over the tex file and then two custom XSLT stylesheets that produce the ODT XML format. I
automated the workflow (generating and zipping) via a Makefile, but it should be relatively simple (and much better) to integrate the workflow into postprocessing (after all that is one of the things Perl is good at; just that I am not good with perl).

I have tested the conversion with a simple test file (source: test.tex/campanile_fog.jpg and result: test.odt attached), it can deal with sectioning, footnotes, images, bibliographies/citations. I am sure that more coverage is relatively easy to add, we just need to understand the ODT format better.

Limitations (after all I only spent around 6 hours on this):

   1.  I do not understand styling very well yet, so I am relying on the standard  styles. Unfortunately, they do not number sections and enumerations. But this can be changed by adding a suitable document format (or changing format in the document)
    2. <ltx:text> is also disregarded until I understand the standard styles better.
    3. figures and captions have not been tackled yet. 
    4. bibtex fields are only covered as far as my simple example needed it. 
    5. it only works with a single bibtex database (in particular, not with embedded bibliography environment)
    6. no bibliography styles for citations; we directly take the reference keys.
    7. the bibliography seems empty at first, but right-click and --> update index table.
    8. I have not even started looking at MathML in ODT yet. But I think that they are using MathML as the formula representation format. 
    9. there should be a way to generate meta.xml with information from the paper frontmatter instead of just copying over an empty one. 
     ...
     99. you will find a lot else that is not covered yet.

I plan to develop LaTeX2ODT further whenever I have a paper to convert. Please feel free to help out.

I chose the ODT format since it is open source. It would probably be just as simple to generate OOXML.

I only just found out that tex4ht does a good job at converting analogously, but it is supposed to be underdocumented and unstable (but probably has more coverage than my converter).