DOCX files are complex, and their complexity makes scraping documents for their content difficult. The aim of this package is to simplify .docx files to just the components which carry meaning, thereby easing the process of pattern matching and data extraction by converting a .docx file into a predictable and human readable JSON file.

Simplifying a complex document down to it's meaningful parts of course requires taking a position on what does and does-not convey meaning in a document. Generally, this package takes the stance that the document structure (body, paragraphs, tables, etc.) are meaningful as is the text itself, whereas text styling (font, font-weight, etc.) is ignored almost entirely, with the exception of paragraph indentation and numbering which is often used to create lists, block quotes, etc. Furthermore, the opinions expressed by this package are explained in the Options section below and can be changed to suite your needs.

import docx
from simplify_docx import simplify

# read in a document 
my_doc = docx.Document("/path/to/my/favorite/file.docx")

# coerce to JSON using the standard options
my_doc_as_json = simplify(my_doc)

# or with non-standard options
my_doc_as_json = simplify(my_doc,{"remove-leading-white-space":False})

This project relies on the python-docx package which can be installed via pip install python-docx. However, as of this writing, if you wish to scrape documents which contain (A) form fields such as drop down lists, checkboxes and text inputs or (B) nested documents (subdocs, altChunks, etc.), you'll need to clone this fork of the python-docx package.

• "friendly-name": (Default = True): Use user-friendly type names such as "table-cell", over standard element names like "CT_Tc"

• "merge-consecutive-text": (Default = True): Sentences and even single words can be represented by multiple text elements. If True, concatenate consecutive text elements into a single text element.

• "ignore-empty-paragraphs": (Default = True): Empty paragraphs are often used for styling purpose and rarely have significance in the meaning of the document.
• "ignore-empty-text": (Default = True): Empty text runs can make an otherwise empty paragraph appear to contain data.
• "remove-leading-white-space": (Default = True): Leading white-space at the start of a paragraph is ocassionaly used for styling purposes and rarely has significance in the interpretation of a document.
• "remove-trailing-white-space": (Default = True): Trailing white-space at the end of a paragraph rarely has significance in the interpretation of a document.
• "flatten-inner-spaces": (Default = False): Collapse multiple space characters between words to a single space.
• "ignore-joiners": (Default = False): Zero width joiner and non-joiner characters are special characters used to create ligatures in displayed text and don't typically convey meaning (at least in alphabet based languages).

• "dumb-quotes": (Default = True): Replace smart quotes with dumb quotes.
• "dumb-hyphens": (Default = True): Replace en-dash, em-dash, figure-dash, horizontal bar, and non-breaking hyphens with ordinary hyphens.
• "dumb-spaces": (Default = True): Replace zero width spaces, hair spaces, thin spaces, punctuation spaces, figure spaces, six per em spaces, four per em spaces, three per em spaces, em spaces, en spaces, em quad spaces, and en quad spaces with ordinary spaces.
• "special-characters-as-text": (Default = True): Coerce special characters into text equivalents according to the following table:

• "symbol-as-text": (Default = True): Special symbols often cary meaning other than the underlying unicode character, especially when the font is a special font such as Wingdings. If True these are included as ordinary text and their font information is omitted.
• "empty-as-text": (Default = False): There are a variety of "Empty" tags such as the <"w:yearLong"> tag which cause the current year to be inserted into the document text. If True, include these as text formatted as "[yearLong]".
• "ignore-left-to-right-mark": (Default = False): Ignore the left-to-right mark, which is not writeable by pythons csv writer.
• "ignore-right-to-left-mark": (Default = False): Ignore the right-to-left mark which is not writeable by pythons csv writer.

Paragraph style markup are one exception to the styling vs. content dichotomy. For example, block quotes are often indicated by indenting whole paragraphs, and Ordered lists, Unordered lists and nesting of lists is often used to divide sections of a document into logical components.

• "include-paragraph-indent": (Default = True): Include the indentation markup on paragraph (CT_P) elements. Indentation is measured in twips
• "include-paragraph-numbering": (Default = True): Include the numbering styles, which are included in the CT_P.pPr.numPr element. The ilvl attribute indicates the level of nesting (zero based index) and the numId attribute refers to a specific numbering style included in the document's internal styles sheet.

• "simplify-dropdown": (Default = True): Include just the selected and default values, the available options, and the name and label attributes in the form element.
• "simplify-textinput": (Default = True): Include just the current and default values, and the name and label attributes in the form element.
• "greedy-text-input": (Default = True): Continue consuming run elements when the text-input has not ended at the end of a paragraph, and the next block level element is also a paragraph. This typically occurs when the user preses the return key while editing a text input field.
• "simplify-checkbox": (Default = True): Include just the current and default values, and the name and label attributes in the form element.
• "use-checkbox-default": (Default = True): If the checkbox has no value attribute (typically because the user has not interacted with it), report the default value as the checkbox value.
• "checkbox-as-text": (Default = False): Coerce the value of the checkbox to text, represented as either "[CheckBox:True]" or "[CheckBox:False]"
• "dropdown-as-text": (Default = False): Coerce the value of the checkbox to text, represented as "[DropDown:<selected value>]"
• "trim-dropdown-options": (Default = True): Remove white-space on the left and right of drop down option items.
• "flatten-generic-field": (Default = True): generic-fields are CT_FldChar runs which are not marked as a drop-down, text-input, or checkbox. These may include special instructions which apply special formatting to a text run (e.g. a hyper link). If True, the contents of generic-fields are included in the normal flow of text

• "flatten-hyperlink": (Default = True): Flatten hyperlinks, including their contents in the flow of normal text.
• "flatten-smartTag": (Default = True): Flatten smartTag elements, including their contents in the flow of normal text.
• "flatten-customXml": (Default = True): Flatten customXml elements, including their contents in the flow of normal text.
• "flatten-simpleField": (Default = True): Flatten simpleField elements, including their contents in the flow of normal text.

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.