Workflow for converting LaTeX into Open Office / MS Word Format

Question

I often have to write up reports based on the analysis of some data. I use R to analyse the data and export tables, figures, and text. This is then included into a LaTeX document either using input or Sweave (see here for details).

However, when I collaborate with others, I sometimes need to provide a document in Open Office / MS Word format.

Question:

Thus, assume the simplest scenario

• I have a LaTeX document with text, tables, and figures
• I need to export this reliably into Open Office or MS Word format: this includes mathematical formulas, table formatting, and quality figures)
• I don't need to go back from MS Word to LaTeX

What is a reliable, efficient, and preferably free process?

Initial Thoughts:

• Export to HTML,
• OOoLatex
• LaTeX - LyX - ODF
• Use odfWeave
• I also just saw this earlier question that deals with some specific issues associated with LaTeX to Open Office conversion
• I've also found this question about LaTeX to Word for resumes

I was hoping that there is an expert out there who has worked out a good system already.

Answer 1

I implemented this for a large R&D lab. We produced several hundred (if not thousand) documents per year, and the LaTeX Users' community there wanted to be able to produce documents using 'tex as well as WYSIWYG software.

The OP was right in that a well-defined workflow is essential. Part of this is the process, but you may also need to think about training and using a common repository, and how to implement corporate design.

Process

We implemented a process that allowed people to work in LaTeX and then switch to .docx for collaborators.

1. Define a class file that contains the correct formatting, etc, using article, report or book classes. Include the minimum number of up-to-date packages in the class and add the nag package to make sure that you (and other users) can see that those packages are not deprecated.
2. Create a template showing how to use the class file
3. Create an SVN (or git, or whatever) repository for the class and template files, and distribute the URL of the repository to LaTeX users
4. Create documents using the lab-standard class file
5. Convert the tex files to .docx using Pandoc, which works on Windows, Mac, and Linux
6. Get edits and peer reviews done on the .docx
7. Transfer edits from the .doc or .docx document back in to 'tex manually, and complete the PDF production in LaTeX.
8. Tagging the document using Adobe Acrobat for Section 508 compliance (accessibility).

N.B. Using one of the web-based editors like sharelatex.com or overleaf.com can remove the need for 5-7, especially now that they have rather good review tools.

Challenges

There were a couple of challenges we had to face to get this adopted.

1. Getting the editors and reviewers something that fit with their existing process, hence the use of the .docx format
2. Figuring out how to get the same class file(s) to all users, hence the SVN repository
3. Making sure people know how to use it, hence the template
4. Figuring out tools that let people collaborate. But that's a whole other post!

508 Compliance / Structured PDFs

The one thing that is still causing trouble is 508-compliance. I have been working (slowly) on using the pdfcomment package to add tooltips and modifying the accessibility package so that documents are accessible. My test PDF documents sometimes pass automated testing in Adobe Acrobat...

Repository

I've put a set of demo documents in a Github repository which may be helpful.

Note re. Pandoc

3 Dec 2017: Originally I suggested the use of latex2rtf instead of Pandoc. I am now editting this answer to suggest Pandoc as I find Pandoc is kept up to date, works well, and I like the flexibility to choose from many more input and output file types.

Answer 2

I think that LaTeX is the wrong starting format, especially if you are generating your input file using Sweave. Instead you can consider using a light-weight markup (Markdown, RST, etc) as a starting format. It will be much easier to convert these formats to both LaTeX and OpenOffice (for example, using Pandoc). As an example, see this sweave file which is written in Markdown. I processed it using sweave, did a bit of post-processing, and then used Pandoc to convert it into ConTeXt. Since the file after post-processing is completely in Markdown format, converting it to OpenOffice should not be a problem.

Answer 3

The new version of Word (2013) lets you open and edit PDFs. The workflow is then:

1. Use latex and pdflatex to make your PDF
2. Open the PDF in Word 2013
3. Save as docx

Answer 4

I think these two softwares are missing in the list.

1. TeX2Word from Chikriilab

2. LaTeX-to-Word from Grindeq

Both of them work elegantly for a properly written LaTeX file. Also, they offer packages for word to LaTeX conversion which are again excellent. But unfortunately both of them are not free.

Answer 5

I found a very easy solution for converting LaTeX-documents into editable Word-files.

1. Compile your LaTeX-document to PDF
2. Go to the Internet-page http://pdftoword.com/
3. Upload your PDF and wait until the Word-file arrives.

I have only tested the site with text-files (no graphics or formulas), but it converted complex contract in Norwegian (æøå) to pretty exact copies. You loose the structure (no styles, only direct formatting), but it works if you need to send a Word-file for proof reading etc.

I suggest setting the text ragged-right in LaTeX. This turns off hyphenation (i.e. do not use ragged2e) and the word document will be easier to edit.

Of course, later you have to merge any changes to your LaTeX-source, but still it is better than retyping the document.

For the sake of good order: I have no connections with Nitro Software, I do not even own a copy of their program.

Answer 6

There is no pain-free way to do this. Really.

Convert your beautiful TeX to pdf, run pdftotext and then import the plain text into a word processor. Recreate all of the tables and equations by hand. Waste days of your life in order to be "compatible" with chumps who don't care about typography until, finally, you decide to stop working with them. Only then will you find inner peace.

Answer 7

My first instinct would probably be oolatex too, or some other technique using TeX4ht, but another method that can also work well is latex2rtf, though I've had the best luck when I tell it it convert formulas, tables, and other complicated stuff to embedded images in the result: obviously, this isn't a great option if the people you're sending them to need to be able to edit those formulas, etc. (But fine if they only need to read and comment.)

Answer 8

Several people have mentioned tex4ht but didn't give the command. From my looking around it seems that the command to run is mk4ht oolatex myfile.tex and you should get a .odt file. I tried it on a basic example and it worked great. When I get a chance I will run it on something more complex.

Answer 9

The best way I know to convert a TeX to an XML application is tex4ht. The project page says it converts TeX to a number of different output formats, including "(X)HTML, MathML, OpenDocument, and DocBook." I believe tex4ht can even convert tikz code to SVG graphics. Word supports OpenDocument, so in theory you could just open up the converted document in Word. I'd expect tables to survive the transition, not so much equations and figures. But MS Word's native format is also an XML application, so you might be able to write an XSLT stylesheet to handle the math and figures.

The need for this kind of tool is evident, and the fact that there's no polished way to do it yet somewhat indicates the complexity of the task. Keep that in mind before you take it up!

Answer 10

Another unlisted solution is the full version of Adobe Acrobat.

I tried the majority of the solutions listed here, which all failed pretty miserably.

Adobe successfully converted nearly everything perfectly, including:

• most equations
• almost all formatting
• images/generated figures

Answer 11

If all else fails (it did for me), Word 2016 can open pdfs directly. It does a decent job of converting them (I was surprised!), although equations are often converted into images.

Answer 12

If you are not forced to stick to a certain format pick your weapon of choice -- tex4ht (you can just use oolatex) , tth, latex2html etc -- and prepare a document style that converts well with that. I do this all the time for simple reports and such that I need to share with people who like using MS-Word etc to edit them. If you spend a bit of time to taylor it for the conversion, you can get consistently good results.

If you have requirements to stick a certain format, for example for a grant proposal etc, you can get by picking up a style that has more or less the right format but with minimal extras to generate the text then use MS-Word or OpenOffice to fix it up.

Answer 13

All answers above suggested to use some converter from tex/pdf file to the wanted file format, that is why I try to give an n-th proposal. I think this approach is quite insane in this situation, as native solutions also exist - as the OP also mentioned.

As you generate the reports from R, it might be the less painful to rewrite some function you use in the reporting process and update those to be able to run in odfWeave. Well, it will generate an odt file from an odt one, so not a native Word format, but it is compatible with Ms Office also from the 2007 version (SP2).

That would require to write the body of your text (if any) and the reporting R code in a word processor (Ms Word or e.g. OOWriter), and later run it via odfWeave. The package has a really great documentation, just download the sources and look for the formatting.odt in the examples directory, which shows in 30+ pages most of the great formatting features of the package. This includes: paragraph, font, color, table, cell, image etc. also.

Answer 14

I don't see Nuance PDF Converter listed in other responses, and I think it is worth a mention. It is not free, but the Mac version that I downloaded has a free trial. I bought it after it successfully converted a PDF that I generated from LaTeX into a clean Word document. Adobe failed miserably on the tables, but Nuance worked well.

Answer 15

I found a good solution through Latex2rtf + TexSword. The process consists performing first the convertion LaTex-> Word (which in my case is around 85% correct), and then to fix the wrong or not converted parts with TexSword (the remaining 15%).

Answer 16

I have struggled for quite some time but with little success. I tried Latex2rtf and it did not convert references, formulas and tables correctly. tex4ht helped me a lot. I used the following steps to successfully convert tex file to doc file with references.

1. install text4ht
2. run latex article.tex
3. run biber article
4. run latex article.tex
5. run mk4ht oolatex article.tex

Now, you can open the newly created file article.odt in OpenOffice (LiberOffice). You can save this file into doc or docx format as desired.

Answer 17

For me there are not doubt: The best workflow is start with Rmarkdown (edit in 2023: or today Quarto) and knitr (instead of Sweave), and then to compile the same simple source file document in Rstudio or VS Code as LaTeX, or PDF (made with a LaTeX compiler), or as HTTML, or as Word document directly.

The bad news is that Rmardownk syntax can produce the fundamental LaTeX constructs (as sections levels, cites, links, figures ... and tables and even a toc) but not every posible LaTeX code.

The good news is that (a) for a Word conversion this is mostly irrelevant because other type of LaTeX constructs often cannot be exported to Word in any way. i.e., there are not a clear Word equivalent (b) statistical reports rarely need other special thing that tables and figures generated by R, i.e., the basic markdown structures plus R chunks are more than enough. And (c) for a special high quality PDF, you can insert LaTeX commands directly in the Rmarkdown/Quarto text (although, unfortunately, will be used only in case of a PDF output). A small example:

---
title: "Test"
author: "Someone"
date: "26 de abril de 2019"
output:
  word_document: default
  pdf_document: default
---
## A test

```{r}
a <- mean(1:10)
```
 The mean is `r a`. This is not the end. 
 \mbox{This is the end with \LaTeX}

(click on the imagen to zoom it)

Answer 18

It sounds like the process you are describing can be entirely offloaded to Authorea. You can write your manuscript in LaTeX, Markdown (and rich text) right on the web and you can even embed data of various sorts (CSV, R files, d3.js, plot,ly to Jupyter Notebooks). Your manuscript has an underlying Git repository. At any point in time, you can export to Word (docx), as well as PDF, or LaTeX - also customizing the look and citation style of your exported manuscript.

Answer 19

The recommended way to convert LaTeX to Office formats using tex4ht is to use make4ht:

make4ht -f odt filename.tex

In contrast to mk4ht oolatex it does lots of post-processing of the generated files, to fix issues with Unicode and other stuff that might result in invalid ODT file otherwise.

The direct ODT output, in contrast to import of HTML to LO or Word, should keep better the structural information, especially math, tables or footnotes.

Answer 20

Some answers already mentioned Adobe Acrobat, but I just want to expand on that a little. I just had to convert a 130 pages thesis from Latex to Word format for hand-in.

Adobe Acrobat Pro DC worked astonishingly well. The export engine has two options: preserve text flow, and preserve page layout.

I used 'preserve text flow'. Some things to note:

• Layout is incredibly well preserved, not messing with the pages apart from a couple of words here or there
• Table of contents is recognised as a TOC, preserving references/hyperlinks as well
• List of figures and tables preserved the references/hyperlinks
• Chapters and (sub)sections are recognised as such
• Itemize/enumerate are recognised as such
• Booktabs are imported as actual tables, keeping layout and style
• Multicolumn environments are recognised as such
• Hyphenation is not recognised as Word hyphenation, but as hyphen followed by space
• Footnotes are not recognised as Word footnotes, but as superscript text
• Longtables are a mess
• Listings with borders and numbered lines are somewhat a mess. Indentation was messed up in some lines and some borders misplaced. Line numbers were sometimes recognised as a numbered list, sometimes just as text.
• PGFplots and ggplots from R are imported as vector graphics, with separate elements. However, some points were hidden.

It is possible, however, to just copy some things over from a 'preserve page layout' version. With this option, text and such is placed within textboxes, thus keeping layout very well. In this mode, PGFplots/ggplots and longtables were displayed accurately. So my workflow was:

1. Export with 'preserve text flow' and 'preserve page layout'
2. Copy figures and longtables from 'page layout' to 'text flow' version
3. As I needed a somewhat accurate word count, I had to replace Latex hyphenation with Word hyphenation to avoid double counts. Search for '- ' and replace with '', manually stepping through to avoid any mistakes. Then turn on hyphenation in Word. Worked not too bad, a couple of words here and there messing up the layout.
4. Check layout page by page, also look for footnotes ending up on the wrong page.
5. Manually tweak listings: indentation, borders and line numbers.

So that's not exactly a perfect automated workflow for reproducible analysis, but it gives pretty decent results even for large projects.

Answer 21

Since I did not come across it here, I am just highlighting that instead of html or PDF as an in-between format, ASCIIDOC (converted from LaTex by Pandoc) appears to be another alternative.

From: http://dag.wiee.rs/home-made/unoconv/#Screenshot

Screenshot

This may not be your typical set of screenshots, but here is a workflow I use for converting an AsciiDoc (text) formatted résumé into professional ODT, PDF, HTML and DOC versions:

[dag@moria cv]$ make odt pdf html doc
rm -f *.{odt,pdf,html,doc}
asciidoc -b docbook -d article -o resume.xml resume.txt
docbook2odf -f --params generate.meta=0 -o resume.tmp.odt resume.xml
# Saved resume.tmp.odt
unoconv -f odt -t template.ott -o resume.odt resume.tmp.odt
unoconv -f pdf -t template.ott -o resume.pdf resume.odt
unoconv -f html -t template.ott -o resume.html resume.odt
unoconv -f doc -t template.ott -o resume.doc resume.odt

The original files are:

A recipe: Makefile  
An AsciiDoc [.txt] file: resume.txt 
An OpenOffice template [.ott] file: template.ott

Converted into the following files:

Open Document [.odt]: resume.odt    - converted to ODF from TXT using asciidoc and stylesheet
PDF [.pdf]: resume.pdf  - converted from ODF using unoconv
HTML [.html]: resume.html   - converted from ODF using unoconv
Word [.doc]: resume.doc - converted from ODF using unoconv 

Answer 22

The best result that I achieved was following the steps:

1. htlatex document.tex;
2. Open document.html in LibreOffice
3. Export as *.odt;
4. Open *.odt and save as *.docx
5. Remove all annotations

The formulas stay as image and references and bibliography are included.

Answer 23

I tried various converters and combinations, but I obtained the best results with the following procedure.

1) Use htlatex to produce HTML code, with the following options:

htlatex document.tex "xhtml,mathml" " -cunihtf -utf8"

2) Convert the html with pandoc:

pandoc -s document.html -o document.docx

This procedure generates a docx with editable equations and a reasonable typesetting. In particular, getting editable equations is not easy and is not always guaranteed by the other methods. The main drawback of my method is that the figures are converted to raster png files, so the quality is degraded. But, typically, who uses Word is not too much interested into such technical details, so none of my colleagues complained about this problem (or even noticed).

However, there can be some difficulties. Some packages do not work properly, so, if any difficulty is found, a careful debugging must be done in order to find and remove the guilty package. Here is a list of issues I found.

1) The .html file does not display properly in my browser, although the docx obtained from it through pandoc is fine.

2) The package mhchem does not produce the desired result. The limited support of htlatex by mhchem is declared in the documentation. I had to write a perl script which rewrites the mhchem commands in an easier way.

3) The \mathrm{something} command does not make "something" upright, in my case. The problem is that the "mathrm" is encoded as mathvariant="normal" by htlatex, which should be fine, but then the "normal" is rendered as italic by pandoc. The workaround is to substitute "normal" with something upright, such as "sans-serif". So, between htlatex and pandoc, I run:

cat document.html | sed 's/mathvariant=\"normal\"/mathvariant=\"sans-serif\"/g' >memo.html

mv memo.html document.html

The drawback is that the mathrm text becomes "sans-serif", not necessarily the font we want.

4) Depending on the version and installation of htlatex, I had problems with the figures included with includegraphics. They are converted from the original format to png by calling some programs, which sometimes do not exist. I was not able to fix the bug, but I found a workaround: I add a cfg file that defines the conversions and pass its name to htlatex. So htlatex uses the programs defined by me instead of using its default converters (which do not exist). It takes time but it works.

Answer 24

I've been revisiting this recently (October 2019).

Originally I focussed on the "how to convert to word" part of the question and I advocated using latex and modifying the output of that for different users. I think that is still valid in some circumstances and will leave that answer.

I came back to this in the context of reproducible research, i.e. linking analysis to publications and preparing different output formats, which is closer to the original question.

I now think the key is to combine the analysis and documentation in the simplest possible base format, run the analysis, and then convert the resulting documentation to whatever output format is required for whichever group. This technology has become dramatically better in the last few years.

The workflow I have settled on is...

1. Prepare an R markdown file that includes analysis steps and documentation. The analysis is included in knitr chunks and can be in R, python, and many other languages.
2. Set up the YAML preamble in the R markdown to describe the output document format(s) I want. In my case they include PDF, HTML, word, and Github-flavoured markdown.
3. Render this in R to produce the different documents. This is done using the rmarkdown and bookdown packages in R. rmarkdown and bookdown leverage pandoc to produce very good output files.
4. Share the different output files with different groups.
5. Update the Rmarkdown and iterate items 1-4 until complete.

I have put an example of this into a repository at https://github.com/AndyClifton/lit-pro-sci-pub-demo.

Answer 25

This is my method for converting LaTeX documents to MS Word documents: first convert LaTeX file to epub3 file with MathML support by using tex4ebook, then convert its output to Word file by using Pandoc.

Conversion from LaTeX to epub3: tex4ebook -f epub3 mwe.tex "mathml"

Conversion from epub3 to MS Word: pandoc mwe.epub -o mwe.docx

Note: Probably you will need to revise your LaTeX source file before converting it to epub3, see this answer for details.

Answer 26

I have had little luck with, e.g. htlatex and mk4ht.

However, what works for me are the following four strategies:

1. pdftotext -layout thesis.pdf
2. detex thesis.tex > thesis_detex.txt
3. pdf2htmlEX thesis.pdf
4. Adobe Acrobat XI Pro

Pdftotext produces thesis.txt, which is very close to the formatting in pdf. To avoid character-mapping problems, load thesis.txt in a windows text editor, e.g., notepad++, select all and paste into a new MS Word/Openoffice Writer document. The only drawback is that if your tex-file produces a pdf with concatenated words, thesis.txt will contain the concatenations.

Detex produces thesis_detex.txt which works fine for many use cases since words are not concatenated.

pdf2htmlEX produces a very accurate html file. Although some characters are collated to one new character.

To be honoust, the conversion with Adobe Acrobat XI Pro worked flawlessly. I was surprised since I had not used Adobe for a while. Earlier versions did not perform well.

In conclusion, the best option for me is to use Adobe on a latex produced pdf.

Answer 27

• Step 1. Create a PDF from your LaTeX file using pdflatex
• Step 2. Convert the newly created PDF to Word document. You can achieve this step in 3 different ways:
• For Windows users, Microsoft 2013 and later versions can open a PDF and save it as Word.
• If you're not using Microsoft Word, you can simply use an online PDF to Word Converter as an alternative, no download needed.
• If your document is text-only, just open it with Google Docs, and then download it as a Word document.