Documentation – 使用 Python 从 PDF 中提取文本

Python-Net: 使用 Python 的基本文本提取

Mon, 08 Jun 2026 00:00:00 +0000

提取 PDF 文档所有页面的文本

使用 TextAbsorber 捕获 PDF 文档每一页的全部文本并将其写入文本文件。此方法非常适合将 PDF 转换为可搜索的文本、进行内容分析，或为索引和后续处理准备文本。

使用以下方法打开 PDF 文档文档.
创建一个 TextAbsorber 实例。
调用 document.pages.accept(text_absorber) 扫描所有页面。
检索提取的文本来自 text_absorber.text.
将结果写入输出文本文件。

import os
import aspose.pdf as ap


def extract_text_from_all_pages(infile, outfile):
    """
    Extract all text from every page of the PDF and write to an output text file.
    Args:
        infile (str): Path to input PDF file.
        outfile (str): Path to output text file.
    """
    # Open the PDF document
    document = ap.Document(infile)
    # Create a TextAbsorber to extract text
    text_absorber = ap.text.TextAbsorber()
    # Accept the absorber for all pages
    document.pages.accept(text_absorber)
    # Get extracted text
    extracted_text = text_absorber.text
    # Write the text to an output file
    with open(outfile, "w", encoding="utf-8") as tw:
        tw.write(extracted_text)

从特定页面提取文本

应用 TextAbsorber 对单个页面进行操作，以隔离并保存多页文档中该部分的文本。这在您只需要单页内容时非常有用——例如，发票、报告章节或表单摘要。

使用以下方法打开 PDF 文档文档.
创建一个 TextAbsorber 实例。
调用 accept 在目标页面上： document.pages[page_number].accept(text_absorber).
检索提取的文本并将其写入文件。

import os
import aspose.pdf as ap


def extract_text_from_page(infile, outfile, page_number):
    """
    Extract text from a specific page number of the PDF.
    Args:
        infile (str): Path to input PDF file.
        outfile (str): Path to output text file.
        page_number (int): 1-based page index to extract.
    """
    document = ap.Document(infile)
    text_absorber = ap.text.TextAbsorber()
    # Accept the absorber on only the specified page
    document.pages[page_number].accept(text_absorber)
    extracted_text = text_absorber.text
    with open(outfile, "w", encoding="utf-8") as tw:
        tw.write(extracted_text)

Python-Net: 使用 Python 的基于区域的提取

Mon, 08 Jun 2026 00:00:00 +0000

从页面的特定区域提取文本

使用 TextAbsorber 与 a 矩形将提取限制在页面的特定区域。此方法对于从页眉、页脚、表格单元格、表单字段、发票或其他已预先知晓文本位置的固定布局区域进行基于区块的提取非常有用。

打开源 PDF 作为文档.
创建一个 TextAbsorber 实例。
配置 text_search_options 将提取限制在矩形内。
在目标页面接受吸收器。
将提取的文本写入输出文件。

import aspose.pdf as ap


def extract_text_from_region(infile, page_number, rect_coords, outfile):
    """
    Extract text from a specified rectangular region on a given page.
    Args:
        infile (str): Path to input PDF file.
        page_number (int): 1-based index of the page.
        rect_coords (tuple): (llx, lly, urx, ury) coordinates of the rectangle.
        outfile (str): Output text file path.
    """
    document = ap.Document(infile)
    try:
        absorber = ap.text.TextAbsorber()
        # Set options to restrict search to the rectangle
        absorber.text_search_options.limit_to_page_bounds = True
        llx, lly, urx, ury = rect_coords
        absorber.text_search_options.rectangle = ap.Rectangle(llx, lly, urx, ury, True)
        # Accept on the specific page
        document.pages[page_number].accept(absorber)
        extracted_text = absorber.text
        with open(outfile, "w", encoding="utf-8") as tw:
            tw.write(extracted_text)
    finally:
        document.close()

通过遍历它们来提取段落

使用 ParagraphAbsorber 当您需要段落感知的提取而不是普通的页面文本时。不同于 TextAbsorber 或 TextFragmentAbsorber, 此 API 按页面、章节和段落组织输出，这对于文本分析、结构化导出以及布局敏感的处理非常有用。

打开源 PDF 作为文档.
创建一个 ParagraphAbsorber 实例。
调用 absorber.visit(document) 分析所有页面。
遍历 page_markups，然后遍历每个章节和段落。
从每个段落读取文本片段，并将结果写入文件。

import aspose.pdf as ap


def extract_paragraphs_from_pdf(infile, outfile):
    """
    Extract all paragraphs from a PDF document, and write each paragraph’s text into an output file.
    Args:
        infile (str): Path to input PDF file.
        outfile (str): Path to output text file.
    """
    document = ap.Document(infile)
    try:
        absorber = ap.text.ParagraphAbsorber()
        absorber.visit(document)

        with open(outfile, "w", encoding="utf-8") as tw:
            for page_markup in absorber.page_markups:
                for sec_idx, section in enumerate(page_markup.sections, start=1):
                    for para_idx, paragraph in enumerate(section.paragraphs, start=1):
                        # Concatenate all fragments/lines in the paragraph
                        parts = []
                        for line in paragraph.lines:
                            for fragment in line:
                                parts.append(fragment.text)
                            parts.append("\r\n")
                        paragraph_text = "".join(parts)
                        tw.write(
                            f"Page {page_markup.number}, Section {sec_idx}, Paragraph {para_idx}:\n"
                        )
                        tw.write(paragraph_text + "\n")
    finally:
        document.close()

提取段落并呈现其边界多边形

您也可以使用 ParagraphAbsorber 以检查段落几何形状。除了提取文本之外，此方法还记录每个章节矩形和段落多边形，这对于布局映射、文档分析、可访问性工具或基于区域的后处理非常有用。

打开源 PDF 作为文档.
创建一个 ParagraphAbsorber 实例。
访问目标页面。
读取页面标记来自 absorber.page_markups.
遍历章节和段落以捕获几何信息和文本。
将矩形、多边形和文本数据写入输出文件。

import aspose.pdf as ap


def extract_paragraphs_with_geometry(infile, outfile):
    """
    Extract paragraphs and record geometry info (rectangle / polygon) for each paragraph in a PDF.
    Args:
        infile (str): Path to input PDF file.
        outfile (str): Path to output text file.
    """
    document = ap.Document(infile)
    try:
        absorber = ap.text.ParagraphAbsorber()
        absorber.visit(document.pages[1])  # Visit page 2 (pages are 1-indexed)

        page_markup = absorber.page_markups[0]
        with open(outfile, "w", encoding="utf-8") as tw:
            for sec_idx, section in enumerate(page_markup.sections, start=1):
                tw.write(f"Section {sec_idx}: rectangle = {section.rectangle}\n")
                for para_idx, paragraph in enumerate(section.paragraphs, start=1):
                    tw.write(f"  Paragraph {para_idx}: polygon = {paragraph.points}\n")
                    # Concatenate paragraph text
                    parts = []
                    for line in paragraph.lines:
                        for fragment in line:
                            parts.append(fragment.text)
                        parts.append("\r\n")
                    tw.write("    Text: " + "".join(parts) + "\n\n")
    finally:
        document.close()

Python-Net: 改进多列 PDF 的文本提取

Mon, 08 Jun 2026 00:00:00 +0000

手动减小字体大小后再提取

在某些多列布局中，提取前减小文本片段的字体大小可以改善阅读顺序并减少重叠问题。此技术有助于处理布局紧凑的文档，如杂志、研究论文、宣传册或包含密集文本列的报告。

加载 PDF。
使用 TextFragmentAbsorber 收集文本片段。
将每个片段的字体大小减小，然后保存并重新打开文档。
使用 TextAbsorber 提取文本。
将提取的文本写入输出文件。

import io
import aspose.pdf as ap


def extract_text_reduce_font(infile, outfile, reduce_ratio=0.7):
    """
    Extract text from a multi-column PDF by first reducing font size of all text fragments.
    Args:
        infile (str): Path to input PDF.
        outfile (str): Output text file.
        reduce_ratio (float): Ratio to reduce font size (e.g., 0.7 = 70%).
    """
    doc = ap.Document(infile)

    frag_absorber = ap.text.TextFragmentAbsorber()
    doc.pages.accept(frag_absorber)
    for frag in frag_absorber.text_fragments:
        frag.text_state.font_size = frag.text_state.font_size * reduce_ratio
    # Save to memory stream and reopen (to apply changes)
    ms = io.BytesIO()
    doc.save(ms)
    ms.seek(0)
    doc2 = ap.Document(ms)
    text_absorber = ap.text.TextAbsorber()
    doc2.pages.accept(text_absorber)
    extracted_text = text_absorber.text
    with open(outfile, "w", encoding="utf-8") as tw:
        tw.write(extracted_text)

使用比例因子提取文本

多列提取的另一种选项是配置 TextExtractionOptions 使用比例因子。调整比例因子可以改善对紧密排列片段的解释，并有助于在密集布局、表格或基于列的文档中保持更准确的阅读顺序。

加载 PDF。
创建一个 TextAbsorber.
配置 TextExtractionOptions.scale_factor.
将提取选项分配给 absorber。
提取页面文本并将结果写入输出文件。

import aspose.pdf as ap


def extract_text_scale_factor(infile, outfile, scale_factor=0.5):
    """
    Extract text from a PDF with multi-column layout using scale factor.
    Args:
        infile (str): Input PDF path.
        outfile (str): Output text file path.
        scale_factor (float): Scale factor between 0.1 and 1.0 or 0 for auto-scaling.
    """
    doc = ap.Document(infile)
    text_absorber = ap.text.TextAbsorber()
    ext_opts = ap.text.TextExtractionOptions(
        ap.text.TextExtractionOptions.TextFormattingMode.PURE
    )
    ext_opts.scale_factor = scale_factor
    text_absorber.extraction_options = ext_opts
    doc.pages.accept(text_absorber)
    extracted_text = text_absorber.text
    with open(outfile, "w", encoding="utf-8") as tw:
        tw.write(extracted_text)

Python-Net: 使用 Python 的批注和特殊文本

Mon, 08 Jun 2026 00:00:00 +0000

从印章批注中提取文本

使用 TextAbsorber 提取嵌入的文本 StampAnnotation 外观流。当印章内容以表单 XObject 的形式渲染而不是以纯文本形式存储时，这非常有用。

打开文档.
访问目标注释 page.annotations.
验证它是一个 StampAnnotation，然后检索其正常外观 XForm。
将 XForm 传递给 TextAbsorber.visit() 提取嵌入的文本。

import os
import aspose.pdf as ap


def extract_text_from_stamp(infile, page_number, annotation_index, outfile):
    """
    Extracts text from a stamp annotation on a given page in a PDF document.
    Args:
        infile (str): Path to the input PDF file.
        page_number (int): 1-based index of the page containing the stamp.
        annotation_index (int): 1-based index of the annotation in that page.
        outfile (str): Path to the output text file where extracted text will be saved.
    """
    document = ap.Document(infile)
    try:
        page = document.pages[page_number]
        annot = page.annotations[annotation_index]
        # Ensure it's a StampAnnotation
        if isinstance(annot, ap.annotations.StampAnnotation):
            # Get normal appearance XForm of the stamp
            xform = annot.appearance["N"]
            absorber = ap.text.TextAbsorber()
            absorber.visit(xform)
            extracted = absorber.text
            with open(outfile, "w", encoding="utf-8") as f:
                f.write(extracted)
    finally:
        document.close()

提取突出显示的文本

遍历页面的注释并使用 HighlightAnnotation.get_marked_text() 读取每个突出显示所覆盖的文本跨度。页面注释集合是从 1 开始的。

打开文档并选择目标页面。
遍历 page.annotations.
使用 is_assignable 用于筛选 HighlightAnnotation 实例。
将注释强制转换后调用 get_marked_text() 检索突出显示的内容。

def extract_highlight_text(infile):
    """
    Extract text from highlight annotations.

    Args:
        infile (str): Input PDF filename

    Returns:
        None

    Example:
        extract_highlight_text("sample.pdf")

    Note:
        Prints marked text from each highlight annotation on first page.
    """
    document = ap.Document(infile)
    page = document.pages[1]

    for annotation in page.annotations:
        if is_assignable(annotation, ap.annotations.HighlightAnnotation):
            highlight_annotation = cast(ap.annotations.HighlightAnnotation, annotation)
            print(highlight_annotation.get_marked_text())

提取上标和下标文本

上标和下标在公式、数学表达式以及化学化合物名称中经常出现。Aspose.PDF for Python via .NET 支持提取这些内容。 TextFragmentAbsorber, 它检测字符级别的定位元数据。

打开文档.
创建一个 TextFragmentAbsorber 实例。
调用 document.pages[page_number].accept(absorber) 扫描目标页面。
从中检索完整提取的文本 absorber.text.
将结果写入文件并关闭文档。

import os
import aspose.pdf as ap


def extract_super_sub_text(infile, outfile, page_number=1):
    """
    Extract text (including superscript/subscript) from a specified page of a PDF and write to a text file.
    Args:
        infile (str): Path to input PDF file.
        outfile (str): Path to output text file.
        page_number (int): 1‑based index of the page to extract.
    """
    document = ap.Document(infile)
    try:
        absorber = ap.text.TextFragmentAbsorber()
        # Accept only the specific page for extraction
        document.pages[page_number].accept(absorber)
        extracted_text = absorber.text
        with open(outfile, "w", encoding="utf-8") as f:
            f.write(extracted_text)
    finally:
        document.close()

遍历文本碎片以检测上标/下标

对于每个片段的检查，遍历 absorber.text_fragments 并阅读 text_state.superscript 和 text_state.subscript 每个的布尔标志 TextFragment.

打开文档并创建一个 TextFragmentAbsorber.
接受目标页上的吸收器以进行填充 absorber.text_fragments.
对于每个片段，读取 fragment.text, fragment.text_state.superscript，和 fragment.text_state.subscript.
将结果写入输出文件并关闭文档。

import os
import aspose.pdf as ap


def extract_super_sub_details(infile, outfile, page_number=1):
    """
    Extract details of each text fragment on a page, identifying superscript and subscript items.
    Args:
        infile (str): Path to input PDF file.
        outfile (str): Path to output text file.
        page_number (int): 1‑based page index.
    """
    document = ap.Document(infile)
    try:
        absorber = ap.text.TextFragmentAbsorber()
        document.pages[page_number].accept(absorber)

        with open(outfile, "w", encoding="utf-8") as f:
            for fragment in absorber.text_fragments:
                text = fragment.text
                is_sup = fragment.text_state.superscript  # True if superscript
                is_sub = fragment.text_state.subscript  # True if subscript
                f.write(
                    f"Text: '{text}' | Superscript: {is_sup} | Subscript: {is_sub}\n"
                )
    finally:
        document.close()