Extracting MS Word document formatting elements along with raw text information

2
votes

In this post @mikemaccana describes how to use python-docx to extract raw text data from an MS Word document from within python. I'd like to go one step further. Instead of simple extracting the raw text information, can I also use this module to harvest information about font face (e.g. bold versus italic) or font size (e.g. 12 versus 18pt). The closest I was able to come was this post asking about using this module to extract highlighted text entries.

Seemed a little abstract, and I'm not totally sure what's going on here. Is there a more straightforward way to extract formatting information from a Word doc in python? By way of a quick document template:

Here the first line is a large header with one sentence.

The second line is slightly smaller. It also has two sentences.

Even smaller. But that's not all. This line has three sentences.

And finally here's a regular line of unbolded text.

If we call these four lines my word document, I'd like to writing a parsing function, call it doc_parser, that returns something like the following: 

>>>> doc_data = doc_parser(path_to_example_doc)
>>>> print(doc_data)
[1] [{'font': 18, 'face': 'bold', 'n_sentence': 1}, 
{'font': 16, 'face': 'bold', 'n_sentence': 2}, 
{'font': 14, 'face': 'bold', 'n_sentence': 3}, 
{'font': 12, 'face': 'plain', 'n_sentence': 1}]
1
pythonms-wordpython-docx

1 Answers

2
votes

The character level formatting ("font") properties are available at the run level. A paragraph is made up of runs. So you can get what you want by going down to that level, like:

for paragraph in document.paragraphs:
    for run in paragraph.runs:
        font = run.font
        is_bold = font.bold
        etc.

The biggest problem you're likely to encounter with that is that the run only knows about formatting that's been directly applied to it. If it looks the way it does because a style has been applied to it, you would have to query the style (which also has a font object) to see what properties it has.

Note that the python-docx that Mike was talking about is the legacy version which was completely rewritten after v0.2.0 (now 0.8.6). Docs are here: http://python-docx.readthedocs.org/en/latest/