Wordファイル(.docx)のテキストとコメントをXMLから読み取るPythonコードを紹介 – python-docx/unstructuredで読み取れない文字も読み取れます！

<w:comment w:initials="KO" w:author="Kaito Osugi" w:date="2024-04-22T12:41:25" w:id="314251227">
    <w:p w:rsidR="63FDECCA" w:rsidRDefault="63FDECCA" w14:paraId="126FF5E0"
        w14:textId="14CD8828">
        <w:pPr>
            <w:pStyle w:val="CommentText" />
        </w:pPr>
        <w:r w:rsidR="63FDECCA">
            <w:rPr />
            <w:t>これはコメントです。</w:t>
        </w:r>
        <w:r>
            <w:rPr>
                <w:rStyle w:val="CommentReference" />
            </w:rPr>
            <w:annotationRef />
        </w:r>
    </w:p>
</w:comment>

import zipfile
from lxml import etree as ET
from typing import Any, List, Optional

namespaces = {"w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main"}
ns = "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}"

def create_comment_id_pair(docx_path: str) -> Optional[dict[str, str]]:
    with zipfile.ZipFile(docx_path, "r") as docx:
        comment_file = "word/comments.xml"
        comment_id = ""
        if comment_file in docx.namelist():
            with docx.open(comment_file) as comments_xml:
                comments_tree = ET.parse(comments_xml, None)
                comments_root = comments_tree.getroot()

                comments_dict = {}
                for comment in comments_root.findall(".//w:comment", namespaces):
                    comment_id = comment.attrib[ns + "id"]
                    comment_text = "".join(
                        elem.text
                        for elem in comment.findall(".//w:t", namespaces)
                        if elem.text
                    )
                    comments_dict[comment_id] = comment_text
                return comments_dict
        else:
            return None

<w:p wp14:paraId="2C078E63"
    xmlns:wp14="http://schemas.microsoft.com/office/word/2010/wordml" wp14:textId="2F3467A2">
    <w:commentRangeStart w:id="314251227" />
    <w:r w:rsidR="150C9252">
        <w:rPr />
        <w:t>私はその人を常に先生と呼んでいた。</w:t>
    </w:r>
    <w:commentRangeEnd w:id="314251227" />
    <w:r>
        <w:rPr>
            <w:rStyle w:val="CommentReference" />
        </w:rPr>
        <w:commentReference w:id="314251227" />
    </w:r>
    <w:r w:rsidR="150C9252">
        <w:rPr />
        <w:t>
            だからここでもただ先生と書くだけで本名は打ち明けない。これは世間を憚かる遠慮というよりも、その方が私にとって自然だからである。私はその人の記憶を呼び起すごとに、すぐ「先生」といいたくなる。筆を執っても心持は同じ事である。よそよそしい頭文字などはとても使う気にならない。私が先生と知り合いになったのは鎌倉である。その時私はまだ若々しい書生であった。</w:t>
    </w:r>
</w:p>

def read_docx_from_xml(
    docx_path: str,
    read_comment: bool = False,
    read_insertion: bool = False,
    read_delete: bool = False,
):
    comment_id_pair = create_comment_id_pair(docx_path)

    with zipfile.ZipFile(docx_path, "r") as docx:
        with docx.open("word/document.xml") as document_xml:
            document_tree = ET.parse(document_xml, None)
            document_root = document_tree.getroot()

        comment_id = ""
        texts: list[str] = []
        for paragraph in document_root.xpath(".//w:p", namespaces=namespaces):
            paragraph_texts: list[Any] = []
            inside_ins = False
            for event, elem in ET.iterwalk(paragraph, events=("start", "end")):
                if read_insertion and event == "start" and elem.tag == ns + "ins":
                    inside_ins = True
                elif read_insertion and event == "end" and elem.tag == ns + "ins":
                    inside_ins = False
                elif event == "end" and elem.tag == ns + "t":
                    if inside_ins:
                        paragraph_texts.append(f"<++{elem.text}++>")
                    else:
                        paragraph_texts.append(elem.text)
                elif read_delete and event == "end" and elem.tag == ns + "delText":
                    paragraph_texts.append(f"<--{elem.text}-->")
                elif (
                    read_comment
                    and event == "start"
                    and elem.tag == ns + "commentRangeStart"
                ):
                    comment_id = elem.attrib.get(ns + "id")
                elif (
                    read_comment
                    and event == "start"
                    and elem.tag == ns + "commentRangeEnd"
                    and elem.attrib.get(ns + "id") == comment_id
                ):
                    if comment_id_pair:
                        comment_text = comment_id_pair.get(comment_id, "")
                        if comment_text:
                            paragraph_texts.append(f"<##{comment_text}##>")

            texts.append("".join(paragraph_texts))

    pretty_texts = []
    for text in texts:
        new_text = (
            text.replace("++><++", "").replace("--><--", "").replace("##><##", "")
        )
        pretty_texts.append(new_text)

    return pretty_texts

    pretty_texts = []
    for text in texts:
        new_text = (
            text.replace("++><++", "").replace("--><--", "").replace("##><##", "")
        )
        pretty_texts.append(new_text)