XML format

OOXML document

The body of a .docx, inside the zip. Text lives in w:t runs, so extracting it means walking runs rather than reading element text.

A .docx is a zip, and word/document.xml is the part that holds the content. Anyone who has tried to read one has met the same surprise: a sentence is not a string. Word splits text into <w:r> runs at every formatting boundary, every spell-check mark and every revision, so "the quick brown fox" may be five runs, and a naive text extraction returns it in pieces or loses the spaces between them.

The other half of the surprise is that the document is not self-contained. Images, hyperlink targets, headers and footnotes live in sibling parts and are referenced by relationship ID through word/_rels/document.xml.rels, so a document part read on its own has dangling references by design.

Constraints that bite

Text is in <w:t>, not in element content generally. Concatenate runs in document order; a run boundary is not a word boundary.
A <w:t> with leading or trailing spaces needs xml:space="preserve", or the space is lost on rewrite.
r:id attributes point into the relationships part, not into this document. Resolving them needs the .rels file from the same zip.
[Content_Types].xml must declare every part's media type, and it is the one part that is not addressable as a relationship.
Element order within a paragraph is significant: <w:pPr> must precede the runs it applies to.

Root element: <document>
Media type: application/xml
Extensions: .xml
Namespaces: http://schemas.openxmlformats.org/wordprocessingml/2006/main
XMLDir label: —

Specification: ECMA-376 Office Open XML

A minimal valid document

Every example on this site is checked against the same parser the workspace uses, so what you see below is known to be well-formed.

ooxml-document.xml

<?xml version="1.0" encoding="UTF-8"?>
<w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main">
  <w:body>
    <w:p>
      <w:pPr><w:pStyle w:val="Heading1"/></w:pPr>
      <w:r><w:t>Quarterly report</w:t></w:r>
    </w:p>
    <w:p>
      <w:r><w:t xml:space="preserve">Revenue grew </w:t></w:r>
      <w:r><w:rPr><w:b/></w:rPr><w:t>12%</w:t></w:r>
      <w:r><w:t> year on year.</w:t></w:r>
    </w:p>
  </w:body>
</w:document>

Validate your own OOXML document Extract its XPath map