处理处理响应

处理请求的响应中会包含一个 Document 对象，其中保存了所处理文档的所有已知信息，包括 Document AI 能够提取的所有结构化信息。

本页面介绍了 Document 对象的布局，提供了一些示例文档，并演示了如何将 OCR 结果的各个方面映射到 Document 对象 JSON 的特定元素。此外，还提供了客户端库代码示例和 Document AI Toolbox SDK 代码示例。 这些代码示例使用在线处理，但 Document 对象解析对于批处理的工作方式相同。

注意：此图表未包含所有可用字段。此图表中的所有 object 字段都有一个或多个具有各自字段的相关对象。

橙色和蓝色矩形及箭头分别表示所连接对象的至少一个字段为 .layout 或 detectedLanguage。该图使用鸦爪表示法。

使用专门用于展开或收起元素的 JSON 查看器或编辑实用程序。在纯文本实用程序中查看原始 JSON 效率低下。

文字、布局和质量得分

以下是一个示例文本文档：

以下是 Enterprise Document OCR 处理器返回的完整文档对象：

下载 JSON

此 OCR 输出也始终包含在 Document AI 处理器输出中，因为 OCR 由处理器运行。它使用现有的 OCR 数据，因此您可以使用内嵌文档选项将此类 JSON 数据输入到 Document AI 处理器中。

  image=None, # all our samples pass this var
  mime_type="application/json",
  inline_document=document_response # pass OCR output to CDE input - undocumented

以下是一些重要字段：

原始文本

text 字段包含 Document AI 识别的文本。此文本不包含除空格、制表符和换行符以外的任何布局结构。这是唯一一个存储文档文本信息并作为文档文本可信来源的字段。其他字段可以按位置（startIndex 和 endIndex）引用文本字段的各个部分。

  {
    text: "Sample Document\nHeading 1\nLorem ipsum dolor sit amet, ..."
  }

页面大小和语言

文档对象中的每个 page 都对应于示例文档中的一个物理页面。示例 JSON 输出包含一个页面，因为输入是单个 PNG 图片。

注意：pages[].pageNumber 从 1 开始，而不是从 0 开始。第一页为第 1 页。

  {
    "pages:" [
      {
        "pageNumber": 1,
        "dimension": {
          "width": 679.0,
          "height": 460.0,
          "unit": "pixels"
        },
      }
    ]
  }

• pages[].detectedLanguages[] 字段包含在给定页面上找到的语言以及置信度分数。

{
  "pages": [
    {
      "detectedLanguages": [
        {
          "confidence": 0.98009938,
          "languageCode": "en"
        },
        {
          "confidence": 0.01990064,
          "languageCode": "und"
        }
      ]
    }
  ]
}

OCR 数据

Document AI OCR 可检测页面中具有各种粒度或组织的文本，例如文本块、段落、词元和符号（符号级别是可选的，如果配置为输出符号级别数据）。这些都是页面对象的成员。

每个元素都有一个对应的 layout，用于描述其位置和文本。非文本视觉元素（例如复选框）也位于页面级别。

{
  "pages": [
    {
      "paragraphs": [
        {
          "layout": {
            "textAnchor": {
              "textSegments": [
                {
                  "endIndex": "16"
                }
              ]
            },
            "confidence": 0.9939527,
            "boundingPoly": {
              "vertices": [ ... ],
              "normalizedVertices": [ ... ]
            },
            "orientation": "PAGE_UP"
          }
        }
      ]
    }
  ]
}

原始文本在 textAnchor 对象中引用，该对象使用 startIndex 和 endIndex 编入主文本字符串。

如果文本字符串从主文本字符串的开头开始，则可以省略

startIndex。

• 对于 boundingPoly，页面的左上角是原点 (0,0)。 X 轴正值表示向右，Y 轴正值表示向下。

• vertices 对象使用与原始图片相同的坐标，而 normalizedVertices 位于 [0,1] 范围内。有一个转换矩阵，用于指示图片归一化的测量倾斜校正和其他属性。

注意：零坐标值会被省略。当 API 检测到坐标（“x”或“y”）值为 0 时，该坐标值将在 JSON 响应中被省略。例如，整个图片周围的 BoundingPoly 所对应的响应将为： "normalizedVertices": [{}, {"x": 1}, {"x": 1,"y": 1}, {"y": 1}]，或 "vertices": [{}, {"x": X_MAX}, {"x": X_MAX,"y": Y_MAX}, {"y": Y_MAX}]

• 如需绘制 boundingPoly，请绘制从一个顶点到下一个顶点的线段。 然后，通过绘制一条从最后一个顶点回到第一个顶点的线段来闭合多边形。布局的 orientation 元素表示文本是否相对于页面进行了旋转。

为了帮助您直观了解文档的结构，以下图片绘制了 page.paragraphs、page.lines、page.tokens 的边界多边形。

段落数

行数

令牌

文本块数

Enterprise Document OCR 处理器可以根据文档的可读性对其进行质量评估。

• 您必须将 processOptions.ocrConfig.enableImageQualityScores 字段设置为 true，才能在 API 响应中获取此数据。

此质量评估是 [0, 1] 中的质量得分，其中 1 表示质量完美。质量得分在 Page.imageQualityScores 字段中返回。所有检测到的缺陷都会列为 quality/defect_*，并按置信度值降序排序。

以下 PDF 文件太暗且模糊，不便于阅读：

下载 PDF 文件

以下是 Enterprise Document OCR 处理器返回的文档质量信息：

  {
    "pages": [
      {
        "imageQualityScores": {
          "qualityScore": 0.7811847,
          "detectedDefects": [
            {
              "type": "quality/defect_document_cutoff",
              "confidence": 1.0
            },
            {
              "type": "quality/defect_glare",
              "confidence": 0.97849524
            },
            {
              "type": "quality/defect_text_cutoff",
              "confidence": 0.5
            }
          ]
        }
      }
    ]
  }

代码示例

以下代码示例演示了如何发送处理请求，然后读取字段并将其输出到终端：

Java

如需了解详情，请参阅 Document AI Java API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

import com.google.cloud.documentai.v1.Document;
import com.google.cloud.documentai.v1.DocumentProcessorServiceClient;
import com.google.cloud.documentai.v1.DocumentProcessorServiceSettings;
import com.google.cloud.documentai.v1.ProcessRequest;
import com.google.cloud.documentai.v1.ProcessResponse;
import com.google.cloud.documentai.v1.RawDocument;
import com.google.protobuf.ByteString;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
import java.util.List;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.TimeoutException;

public class ProcessOcrDocument {
  public static void processOcrDocument()
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // TODO(developer): Replace these variables before running the sample.
    String projectId = "your-project-id";
    String location = "your-project-location"; // Format is "us" or "eu".
    String processerId = "your-processor-id";
    String filePath = "path/to/input/file.pdf";
    processOcrDocument(projectId, location, processerId, filePath);
  }

  public static void processOcrDocument(
      String projectId, String location, String processorId, String filePath)
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // Initialize client that will be used to send requests. This client only needs
    // to be created
    // once, and can be reused for multiple requests. After completing all of your
    // requests, call
    // the "close" method on the client to safely clean up any remaining background
    // resources.
    String endpoint = String.format("%s-documentai.googleapis.com:443", location);
    DocumentProcessorServiceSettings settings =
        DocumentProcessorServiceSettings.newBuilder().setEndpoint(endpoint).build();
    try (DocumentProcessorServiceClient client = DocumentProcessorServiceClient.create(settings)) {
      // The full resource name of the processor, e.g.:
      // projects/project-id/locations/location/processor/processor-id
      // You must create new processors in the Cloud Console first
      String name =
          String.format("projects/%s/locations/%s/processors/%s", projectId, location, processorId);

      // Read the file.
      byte[] imageFileData = Files.readAllBytes(Paths.get(filePath));

      // Convert the image data to a Buffer and base64 encode it.
      ByteString content = ByteString.copyFrom(imageFileData);

      RawDocument document =
          RawDocument.newBuilder().setContent(content).setMimeType("application/pdf").build();

      // Configure the process request.
      ProcessRequest request =
          ProcessRequest.newBuilder().setName(name).setRawDocument(document).build();

      // Recognizes text entities in the PDF document
      ProcessResponse result = client.processDocument(request);
      Document documentResponse = result.getDocument();

      System.out.println("Document processing complete.");

      // Read the text recognition output from the processor
      // For a full list of Document object attributes,
      // please reference this page:
      // https://googleapis.dev/java/google-cloud-document-ai/latest/index.html

      // Get all of the document text as one big string
      String text = documentResponse.getText();
      System.out.printf("Full document text: '%s'\n", escapeNewlines(text));

      // Read the text recognition output from the processor
      List<Document.Page> pages = documentResponse.getPagesList();
      System.out.printf("There are %s page(s) in this document.\n", pages.size());

      for (Document.Page page : pages) {
        System.out.printf("Page %d:\n", page.getPageNumber());
        printPageDimensions(page.getDimension());
        printDetectedLanguages(page.getDetectedLanguagesList());
        printParagraphs(page.getParagraphsList(), text);
        printBlocks(page.getBlocksList(), text);
        printLines(page.getLinesList(), text);
        printTokens(page.getTokensList(), text);
      }
    }
  }

  private static void printPageDimensions(Document.Page.Dimension dimension) {
    String unit = dimension.getUnit();
    System.out.printf("    Width: %.1f %s\n", dimension.getWidth(), unit);
    System.out.printf("    Height: %.1f %s\n", dimension.getHeight(), unit);
  }

  private static void printDetectedLanguages(
      List<Document.Page.DetectedLanguage> detectedLangauges) {
    System.out.println("    Detected languages:");
    for (Document.Page.DetectedLanguage detectedLanguage : detectedLangauges) {
      String languageCode = detectedLanguage.getLanguageCode();
      float confidence = detectedLanguage.getConfidence();
      System.out.printf("        %s (%.2f%%)\n", languageCode, confidence * 100.0);
    }
  }

  private static void printParagraphs(List<Document.Page.Paragraph> paragraphs, String text) {
    System.out.printf("    %d paragraphs detected:\n", paragraphs.size());
    Document.Page.Paragraph firstParagraph = paragraphs.get(0);
    String firstParagraphText = getLayoutText(firstParagraph.getLayout().getTextAnchor(), text);
    System.out.printf("        First paragraph text: %s\n", escapeNewlines(firstParagraphText));
    Document.Page.Paragraph lastParagraph = paragraphs.get(paragraphs.size() - 1);
    String lastParagraphText = getLayoutText(lastParagraph.getLayout().getTextAnchor(), text);
    System.out.printf("        Last paragraph text: %s\n", escapeNewlines(lastParagraphText));
  }

  private static void printBlocks(List<Document.Page.Block> blocks, String text) {
    System.out.printf("    %d blocks detected:\n", blocks.size());
    Document.Page.Block firstBlock = blocks.get(0);
    String firstBlockText = getLayoutText(firstBlock.getLayout().getTextAnchor(), text);
    System.out.printf("        First block text: %s\n", escapeNewlines(firstBlockText));
    Document.Page.Block lastBlock = blocks.get(blocks.size() - 1);
    String lastBlockText = getLayoutText(lastBlock.getLayout().getTextAnchor(), text);
    System.out.printf("        Last block text: %s\n", escapeNewlines(lastBlockText));
  }

  private static void printLines(List<Document.Page.Line> lines, String text) {
    System.out.printf("    %d lines detected:\n", lines.size());
    Document.Page.Line firstLine = lines.get(0);
    String firstLineText = getLayoutText(firstLine.getLayout().getTextAnchor(), text);
    System.out.printf("        First line text: %s\n", escapeNewlines(firstLineText));
    Document.Page.Line lastLine = lines.get(lines.size() - 1);
    String lastLineText = getLayoutText(lastLine.getLayout().getTextAnchor(), text);
    System.out.printf("        Last line text: %s\n", escapeNewlines(lastLineText));
  }

  private static void printTokens(List<Document.Page.Token> tokens, String text) {
    System.out.printf("    %d tokens detected:\n", tokens.size());
    Document.Page.Token firstToken = tokens.get(0);
    String firstTokenText = getLayoutText(firstToken.getLayout().getTextAnchor(), text);
    System.out.printf("        First token text: %s\n", escapeNewlines(firstTokenText));
    Document.Page.Token lastToken = tokens.get(tokens.size() - 1);
    String lastTokenText = getLayoutText(lastToken.getLayout().getTextAnchor(), text);
    System.out.printf("        Last token text: %s\n", escapeNewlines(lastTokenText));
  }

  // Extract shards from the text field
  private static String getLayoutText(Document.TextAnchor textAnchor, String text) {
    if (textAnchor.getTextSegmentsList().size() > 0) {
      int startIdx = (int) textAnchor.getTextSegments(0).getStartIndex();
      int endIdx = (int) textAnchor.getTextSegments(0).getEndIndex();
      return text.substring(startIdx, endIdx);
    }
    return "[NO TEXT]";
  }

  private static String escapeNewlines(String s) {
    return s.replace("\n", "\\n").replace("\r", "\\r");
  }
}

Node.js

如需了解详情，请参阅 Document AI Node.js API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

/**
 * TODO(developer): Uncomment these variables before running the sample.
 */
// const projectId = 'YOUR_PROJECT_ID';
// const location = 'YOUR_PROJECT_LOCATION'; // Format is 'us' or 'eu'
// const processorId = 'YOUR_PROCESSOR_ID'; // Create processor in Cloud Console
// const filePath = '/path/to/local/pdf';

const {DocumentProcessorServiceClient} =
  require('@google-cloud/documentai').v1beta3;

// Instantiates a client
const client = new DocumentProcessorServiceClient();

async function processDocument() {
  // The full resource name of the processor, e.g.:
  // projects/project-id/locations/location/processor/processor-id
  // You must create new processors in the Cloud Console first
  const name = `projects/${projectId}/locations/${location}/processors/${processorId}`;

  // Read the file into memory.
  const fs = require('fs').promises;
  const imageFile = await fs.readFile(filePath);

  // Convert the image data to a Buffer and base64 encode it.
  const encodedImage = Buffer.from(imageFile).toString('base64');

  const request = {
    name,
    rawDocument: {
      content: encodedImage,
      mimeType: 'application/pdf',
    },
  };

  // Recognizes text entities in the PDF document
  const [result] = await client.processDocument(request);

  console.log('Document processing complete.');

  // Read the text recognition output from the processor
  // For a full list of Document object attributes,
  // please reference this page: https://googleapis.dev/nodejs/documentai/latest/index.html
  const {document} = result;
  const {text} = document;

  // Read the text recognition output from the processor
  console.log(`Full document text: ${JSON.stringify(text)}`);
  console.log(`There are ${document.pages.length} page(s) in this document.`);
  for (const page of document.pages) {
    console.log(`Page ${page.pageNumber}`);
    printPageDimensions(page.dimension);
    printDetectedLanguages(page.detectedLanguages);
    printParagraphs(page.paragraphs, text);
    printBlocks(page.blocks, text);
    printLines(page.lines, text);
    printTokens(page.tokens, text);
  }
}

const printPageDimensions = dimension => {
  console.log(`    Width: ${dimension.width}`);
  console.log(`    Height: ${dimension.height}`);
};

const printDetectedLanguages = detectedLanguages => {
  console.log('    Detected languages:');
  for (const lang of detectedLanguages) {
    const code = lang.languageCode;
    const confPercent = lang.confidence * 100;
    console.log(`        ${code} (${confPercent.toFixed(2)}% confidence)`);
  }
};

const printParagraphs = (paragraphs, text) => {
  console.log(`    ${paragraphs.length} paragraphs detected:`);
  const firstParagraphText = getText(paragraphs[0].layout.textAnchor, text);
  console.log(
    `        First paragraph text: ${JSON.stringify(firstParagraphText)}`
  );
  const lastParagraphText = getText(
    paragraphs[paragraphs.length - 1].layout.textAnchor,
    text
  );
  console.log(
    `        Last paragraph text: ${JSON.stringify(lastParagraphText)}`
  );
};

const printBlocks = (blocks, text) => {
  console.log(`    ${blocks.length} blocks detected:`);
  const firstBlockText = getText(blocks[0].layout.textAnchor, text);
  console.log(`        First block text: ${JSON.stringify(firstBlockText)}`);
  const lastBlockText = getText(
    blocks[blocks.length - 1].layout.textAnchor,
    text
  );
  console.log(`        Last block text: ${JSON.stringify(lastBlockText)}`);
};

const printLines = (lines, text) => {
  console.log(`    ${lines.length} lines detected:`);
  const firstLineText = getText(lines[0].layout.textAnchor, text);
  console.log(`        First line text: ${JSON.stringify(firstLineText)}`);
  const lastLineText = getText(
    lines[lines.length - 1].layout.textAnchor,
    text
  );
  console.log(`        Last line text: ${JSON.stringify(lastLineText)}`);
};

const printTokens = (tokens, text) => {
  console.log(`    ${tokens.length} tokens detected:`);
  const firstTokenText = getText(tokens[0].layout.textAnchor, text);
  console.log(`        First token text: ${JSON.stringify(firstTokenText)}`);
  const firstTokenBreakType = tokens[0].detectedBreak.type;
  console.log(`        First token break type: ${firstTokenBreakType}`);
  const lastTokenText = getText(
    tokens[tokens.length - 1].layout.textAnchor,
    text
  );
  console.log(`        Last token text: ${JSON.stringify(lastTokenText)}`);
  const lastTokenBreakType = tokens[tokens.length - 1].detectedBreak.type;
  console.log(`        Last token break type: ${lastTokenBreakType}`);
};

// Extract shards from the text field
const getText = (textAnchor, text) => {
  if (!textAnchor.textSegments || textAnchor.textSegments.length === 0) {
    return '';
  }

  // First shard in document doesn't have startIndex property
  const startIndex = textAnchor.textSegments[0].startIndex || 0;
  const endIndex = textAnchor.textSegments[0].endIndex;

  return text.substring(startIndex, endIndex);
};

Python

如需了解详情，请参阅 Document AI Python API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

from typing import Optional, Sequence

from google.api_core.client_options import ClientOptions
from google.cloud import documentai

# TODO(developer): Uncomment these variables before running the sample.
# project_id = "YOUR_PROJECT_ID"
# location = "YOUR_PROCESSOR_LOCATION" # Format is "us" or "eu"
# processor_id = "YOUR_PROCESSOR_ID" # Create processor before running sample
# processor_version = "rc" # Refer to https://cloud.google.com/document-ai/docs/manage-processor-versions for more information
# file_path = "/path/to/local/pdf"
# mime_type = "application/pdf" # Refer to https://cloud.google.com/document-ai/docs/file-types for supported file types


def process_document_ocr_sample(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
) -> None:
    # Optional: Additional configurations for Document OCR Processor.
    # For more information: https://cloud.google.com/document-ai/docs/enterprise-document-ocr
    process_options = documentai.ProcessOptions(
        ocr_config=documentai.OcrConfig(
            enable_native_pdf_parsing=True,
            enable_image_quality_scores=True,
            enable_symbol=True,
            # OCR Add Ons https://cloud.google.com/document-ai/docs/ocr-add-ons
            premium_features=documentai.OcrConfig.PremiumFeatures(
                compute_style_info=True,
                enable_math_ocr=False,  # Enable to use Math OCR Model
                enable_selection_mark_detection=True,
            ),
        )
    )
    # Online processing request to Document AI
    document = process_document(
        project_id,
        location,
        processor_id,
        processor_version,
        file_path,
        mime_type,
        process_options=process_options,
    )

    text = document.text
    print(f"Full document text: {text}\n")
    print(f"There are {len(document.pages)} page(s) in this document.\n")

    for page in document.pages:
        print(f"Page {page.page_number}:")
        print_page_dimensions(page.dimension)
        print_detected_languages(page.detected_languages)

        print_blocks(page.blocks, text)
        print_paragraphs(page.paragraphs, text)
        print_lines(page.lines, text)
        print_tokens(page.tokens, text)

        if page.symbols:
            print_symbols(page.symbols, text)

        if page.image_quality_scores:
            print_image_quality_scores(page.image_quality_scores)

        if page.visual_elements:
            print_visual_elements(page.visual_elements, text)


def print_page_dimensions(dimension: documentai.Document.Page.Dimension) -> None:
    print(f"    Width: {str(dimension.width)}")
    print(f"    Height: {str(dimension.height)}")


def print_detected_languages(
    detected_languages: Sequence[documentai.Document.Page.DetectedLanguage],
) -> None:
    print("    Detected languages:")
    for lang in detected_languages:
        print(f"        {lang.language_code} ({lang.confidence:.1%} confidence)")


def print_blocks(blocks: Sequence[documentai.Document.Page.Block], text: str) -> None:
    print(f"    {len(blocks)} blocks detected:")
    first_block_text = layout_to_text(blocks[0].layout, text)
    print(f"        First text block: {repr(first_block_text)}")
    last_block_text = layout_to_text(blocks[-1].layout, text)
    print(f"        Last text block: {repr(last_block_text)}")


def print_paragraphs(
    paragraphs: Sequence[documentai.Document.Page.Paragraph], text: str
) -> None:
    print(f"    {len(paragraphs)} paragraphs detected:")
    first_paragraph_text = layout_to_text(paragraphs[0].layout, text)
    print(f"        First paragraph text: {repr(first_paragraph_text)}")
    last_paragraph_text = layout_to_text(paragraphs[-1].layout, text)
    print(f"        Last paragraph text: {repr(last_paragraph_text)}")


def print_lines(lines: Sequence[documentai.Document.Page.Line], text: str) -> None:
    print(f"    {len(lines)} lines detected:")
    first_line_text = layout_to_text(lines[0].layout, text)
    print(f"        First line text: {repr(first_line_text)}")
    last_line_text = layout_to_text(lines[-1].layout, text)
    print(f"        Last line text: {repr(last_line_text)}")


def print_tokens(tokens: Sequence[documentai.Document.Page.Token], text: str) -> None:
    print(f"    {len(tokens)} tokens detected:")
    first_token_text = layout_to_text(tokens[0].layout, text)
    first_token_break_type = tokens[0].detected_break.type_.name
    print(f"        First token text: {repr(first_token_text)}")
    print(f"        First token break type: {repr(first_token_break_type)}")
    if tokens[0].style_info:
        print_style_info(tokens[0].style_info)

    last_token_text = layout_to_text(tokens[-1].layout, text)
    last_token_break_type = tokens[-1].detected_break.type_.name
    print(f"        Last token text: {repr(last_token_text)}")
    print(f"        Last token break type: {repr(last_token_break_type)}")
    if tokens[-1].style_info:
        print_style_info(tokens[-1].style_info)


def print_symbols(
    symbols: Sequence[documentai.Document.Page.Symbol], text: str
) -> None:
    print(f"    {len(symbols)} symbols detected:")
    first_symbol_text = layout_to_text(symbols[0].layout, text)
    print(f"        First symbol text: {repr(first_symbol_text)}")
    last_symbol_text = layout_to_text(symbols[-1].layout, text)
    print(f"        Last symbol text: {repr(last_symbol_text)}")


def print_image_quality_scores(
    image_quality_scores: documentai.Document.Page.ImageQualityScores,
) -> None:
    print(f"    Quality score: {image_quality_scores.quality_score:.1%}")
    print("    Detected defects:")

    for detected_defect in image_quality_scores.detected_defects:
        print(f"        {detected_defect.type_}: {detected_defect.confidence:.1%}")


def print_style_info(style_info: documentai.Document.Page.Token.StyleInfo) -> None:
    """
    Only supported in version `pretrained-ocr-v2.0-2023-06-02`
    """
    print(f"           Font Size: {style_info.font_size}pt")
    print(f"           Font Type: {style_info.font_type}")
    print(f"           Bold: {style_info.bold}")
    print(f"           Italic: {style_info.italic}")
    print(f"           Underlined: {style_info.underlined}")
    print(f"           Handwritten: {style_info.handwritten}")
    print(
        f"           Text Color (RGBa): {style_info.text_color.red}, {style_info.text_color.green}, {style_info.text_color.blue}, {style_info.text_color.alpha}"
    )


def print_visual_elements(
    visual_elements: Sequence[documentai.Document.Page.VisualElement], text: str
) -> None:
    """
    Only supported in version `pretrained-ocr-v2.0-2023-06-02`
    """
    checkboxes = [x for x in visual_elements if "checkbox" in x.type]
    math_symbols = [x for x in visual_elements if x.type == "math_formula"]

    if checkboxes:
        print(f"    {len(checkboxes)} checkboxes detected:")
        print(f"        First checkbox: {repr(checkboxes[0].type)}")
        print(f"        Last checkbox: {repr(checkboxes[-1].type)}")

    if math_symbols:
        print(f"    {len(math_symbols)} math symbols detected:")
        first_math_symbol_text = layout_to_text(math_symbols[0].layout, text)
        print(f"        First math symbol: {repr(first_math_symbol_text)}")




def process_document(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
    process_options: Optional[documentai.ProcessOptions] = None,
) -> documentai.Document:
    # You must set the `api_endpoint` if you use a location other than "us".
    client = documentai.DocumentProcessorServiceClient(
        client_options=ClientOptions(
            api_endpoint=f"{location}-documentai.googleapis.com"
        )
    )

    # The full resource name of the processor version, e.g.:
    # `projects/{project_id}/locations/{location}/processors/{processor_id}/processorVersions/{processor_version_id}`
    # You must create a processor before running this sample.
    name = client.processor_version_path(
        project_id, location, processor_id, processor_version
    )

    # Read the file into memory
    with open(file_path, "rb") as image:
        image_content = image.read()

    # Configure the process request
    request = documentai.ProcessRequest(
        name=name,
        raw_document=documentai.RawDocument(content=image_content, mime_type=mime_type),
        # Only supported for Document OCR processor
        process_options=process_options,
    )

    result = client.process_document(request=request)

    # For a full list of `Document` object attributes, reference this page:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1/Document
    return result.document




def layout_to_text(layout: documentai.Document.Page.Layout, text: str) -> str:
    """
    Document AI identifies text in different parts of the document by their
    offsets in the entirety of the document"s text. This function converts
    offsets to a string.
    """
    # If a text segment spans several lines, it will
    # be stored in different text segments.
    return "".join(
        text[int(segment.start_index) : int(segment.end_index)]
        for segment in layout.text_anchor.text_segments
    )

表单和表格

以下是我们的示例表单：

以下是表单解析器返回的完整文档对象：

下载 JSON

以下是一些重要字段：

表单解析器能够检测页面中的 FormFields。每个表单字段都有一个名称和一个值。这些也称为键值对 (KVP)。请注意，KVP 与其他提取器中的（架构）实体不同：

实体名称已配置。键值对中的键就是文档中的键文本。

{
  "pages:" [
    {
      "formFields": [
        {
          "fieldName": { ... },
          "fieldValue": { ... }
        }
      ]
    }
  ]
}

• Document AI 还可以检测页面中的 Tables。

{
  "pages:" [
    {
      "tables": [
        {
          "layout": { ... },
          "headerRows": [
            {
              "cells": [
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                },
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                }
              ]
            }
          ],
          "bodyRows": [
            {
              "cells": [
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                },
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                }
              ]
            }
          ]
        }
      ]
    }
  ]
}

表单解析器中的表格提取功能只能识别常规表格，即不包含跨行或跨列单元格的表格。因此，rowSpan 和 colSpan 始终为 1。

• 从处理器版本 pretrained-form-parser-v2.0-2022-11-10 开始，表单解析器还可以识别通用实体。如需了解详情，请参阅表单解析器。

• 为了帮助您直观了解文档的结构，下图绘制了 page.formFields 和 page.tables 的边界多边形。

• 表格中的复选框。表单解析器能够将图片和 PDF 中的复选框数字化为 KVP。提供一个以键值对形式进行复选框数字化的示例。

在表格之外，复选框在表单解析器中表示为视觉元素。突出显示界面上带有对勾标记的方框以及 JSON 中的 Unicode ✓。

"pages:" [
    {
      "tables": [
        {
          "layout": { ... },
          "headerRows": [
            {
              "cells": [
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                },
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                }
              ]
            }
          ],
          "bodyRows": [
            {
              "cells": [
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                },
                {
                  "layout": { ... },
                  "rowSpan": 1,
                  "colSpan": 1
                }
              ]
            }
          ]
        }
      ]
    }
  ]
}

在表格中，复选框会显示为 Unicode 字符，例如 ✓（已选中）或 ☐（未选中）。

已勾选的复选框的值为 filled_checkbox：under pages > x > formFields > x > fieldValue > valueType.。未选中的复选框的值为 unfilled_checkbox。

内容字段在路径 pages>formFields>x>fieldValue>textAnchor>content 处将复选框内容值显示为突出显示的 ✓。

为了帮助您直观了解文档的结构，下图绘制了 page.formFields 和 page.tables 的边界多边形。

表单字段

表

代码示例

以下代码示例演示了如何发送处理请求，然后读取字段并将其输出到终端：

Java

如需了解详情，请参阅 Document AI Java API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

import com.google.cloud.documentai.v1.Document;
import com.google.cloud.documentai.v1.DocumentProcessorServiceClient;
import com.google.cloud.documentai.v1.DocumentProcessorServiceSettings;
import com.google.cloud.documentai.v1.ProcessRequest;
import com.google.cloud.documentai.v1.ProcessResponse;
import com.google.cloud.documentai.v1.RawDocument;
import com.google.protobuf.ByteString;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
import java.util.List;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.TimeoutException;

public class ProcessFormDocument {
  public static void processFormDocument()
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // TODO(developer): Replace these variables before running the sample.
    String projectId = "your-project-id";
    String location = "your-project-location"; // Format is "us" or "eu".
    String processerId = "your-processor-id";
    String filePath = "path/to/input/file.pdf";
    processFormDocument(projectId, location, processerId, filePath);
  }

  public static void processFormDocument(
      String projectId, String location, String processorId, String filePath)
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // Initialize client that will be used to send requests. This client only needs
    // to be created
    // once, and can be reused for multiple requests. After completing all of your
    // requests, call
    // the "close" method on the client to safely clean up any remaining background
    // resources.
    String endpoint = String.format("%s-documentai.googleapis.com:443", location);
    DocumentProcessorServiceSettings settings =
        DocumentProcessorServiceSettings.newBuilder().setEndpoint(endpoint).build();
    try (DocumentProcessorServiceClient client = DocumentProcessorServiceClient.create(settings)) {
      // The full resource name of the processor, e.g.:
      // projects/project-id/locations/location/processor/processor-id
      // You must create new processors in the Cloud Console first
      String name =
          String.format("projects/%s/locations/%s/processors/%s", projectId, location, processorId);

      // Read the file.
      byte[] imageFileData = Files.readAllBytes(Paths.get(filePath));

      // Convert the image data to a Buffer and base64 encode it.
      ByteString content = ByteString.copyFrom(imageFileData);

      RawDocument document =
          RawDocument.newBuilder().setContent(content).setMimeType("application/pdf").build();

      // Configure the process request.
      ProcessRequest request =
          ProcessRequest.newBuilder().setName(name).setRawDocument(document).build();

      // Recognizes text entities in the PDF document
      ProcessResponse result = client.processDocument(request);
      Document documentResponse = result.getDocument();

      System.out.println("Document processing complete.");

      // Read the text recognition output from the processor
      // For a full list of Document object attributes,
      // please reference this page:
      // https://googleapis.dev/java/google-cloud-document-ai/latest/index.html

      // Get all of the document text as one big string
      String text = documentResponse.getText();
      System.out.printf("Full document text: '%s'\n", removeNewlines(text));

      // Read the text recognition output from the processor
      List<Document.Page> pages = documentResponse.getPagesList();
      System.out.printf("There are %s page(s) in this document.\n", pages.size());

      for (Document.Page page : pages) {
        System.out.printf("\n\n**** Page %d ****\n", page.getPageNumber());

        List<Document.Page.Table> tables = page.getTablesList();
        System.out.printf("Found %d table(s):\n", tables.size());
        for (Document.Page.Table table : tables) {
          printTableInfo(table, text);
        }

        List<Document.Page.FormField> formFields = page.getFormFieldsList();
        System.out.printf("Found %d form fields:\n", formFields.size());
        for (Document.Page.FormField formField : formFields) {
          String fieldName = getLayoutText(formField.getFieldName().getTextAnchor(), text);
          String fieldValue = getLayoutText(formField.getFieldValue().getTextAnchor(), text);
          System.out.printf(
              "    * '%s': '%s'\n", removeNewlines(fieldName), removeNewlines(fieldValue));
        }
      }
    }
  }

  private static void printTableInfo(Document.Page.Table table, String text) {
    Document.Page.Table.TableRow firstBodyRow = table.getBodyRows(0);
    int columnCount = firstBodyRow.getCellsCount();
    System.out.printf(
        "    Table with %d columns and %d rows:\n", columnCount, table.getBodyRowsCount());

    Document.Page.Table.TableRow headerRow = table.getHeaderRows(0);
    StringBuilder headerRowText = new StringBuilder();
    for (Document.Page.Table.TableCell cell : headerRow.getCellsList()) {
      String columnName = getLayoutText(cell.getLayout().getTextAnchor(), text);
      headerRowText.append(String.format("%s | ", removeNewlines(columnName)));
    }
    headerRowText.setLength(headerRowText.length() - 3);
    System.out.printf("        Collumns: %s\n", headerRowText.toString());

    StringBuilder firstRowText = new StringBuilder();
    for (Document.Page.Table.TableCell cell : firstBodyRow.getCellsList()) {
      String cellText = getLayoutText(cell.getLayout().getTextAnchor(), text);
      firstRowText.append(String.format("%s | ", removeNewlines(cellText)));
    }
    firstRowText.setLength(firstRowText.length() - 3);
    System.out.printf("        First row data: %s\n", firstRowText.toString());
  }

  // Extract shards from the text field
  private static String getLayoutText(Document.TextAnchor textAnchor, String text) {
    if (textAnchor.getTextSegmentsList().size() > 0) {
      int startIdx = (int) textAnchor.getTextSegments(0).getStartIndex();
      int endIdx = (int) textAnchor.getTextSegments(0).getEndIndex();
      return text.substring(startIdx, endIdx);
    }
    return "[NO TEXT]";
  }

  private static String removeNewlines(String s) {
    return s.replace("\n", "").replace("\r", "");
  }
}

Node.js

如需了解详情，请参阅 Document AI Node.js API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

/**
 * TODO(developer): Uncomment these variables before running the sample.
 */
// const projectId = 'YOUR_PROJECT_ID';
// const location = 'YOUR_PROJECT_LOCATION'; // Format is 'us' or 'eu'
// const processorId = 'YOUR_PROCESSOR_ID'; // Create processor in Cloud Console
// const filePath = '/path/to/local/pdf';

const {DocumentProcessorServiceClient} =
  require('@google-cloud/documentai').v1beta3;

// Instantiates a client
const client = new DocumentProcessorServiceClient();

async function processDocument() {
  // The full resource name of the processor, e.g.:
  // projects/project-id/locations/location/processor/processor-id
  // You must create new processors in the Cloud Console first
  const name = `projects/${projectId}/locations/${location}/processors/${processorId}`;

  // Read the file into memory.
  const fs = require('fs').promises;
  const imageFile = await fs.readFile(filePath);

  // Convert the image data to a Buffer and base64 encode it.
  const encodedImage = Buffer.from(imageFile).toString('base64');

  const request = {
    name,
    rawDocument: {
      content: encodedImage,
      mimeType: 'application/pdf',
    },
  };

  // Recognizes text entities in the PDF document
  const [result] = await client.processDocument(request);

  console.log('Document processing complete.');

  // Read the table and form fields output from the processor
  // The form processor also contains OCR data. For more information
  // on how to parse OCR data please see the OCR sample.
  // For a full list of Document object attributes,
  // please reference this page: https://googleapis.dev/nodejs/documentai/latest/index.html
  const {document} = result;
  const {text} = document;
  console.log(`Full document text: ${JSON.stringify(text)}`);
  console.log(`There are ${document.pages.length} page(s) in this document.`);

  for (const page of document.pages) {
    console.log(`\n\n**** Page ${page.pageNumber} ****`);

    console.log(`Found ${page.tables.length} table(s):`);
    for (const table of page.tables) {
      const numCollumns = table.headerRows[0].cells.length;
      const numRows = table.bodyRows.length;
      console.log(`Table with ${numCollumns} columns and ${numRows} rows:`);
      printTableInfo(table, text);
    }
    console.log(`Found ${page.formFields.length} form field(s):`);
    for (const field of page.formFields) {
      const fieldName = getText(field.fieldName.textAnchor, text);
      const fieldValue = getText(field.fieldValue.textAnchor, text);
      console.log(
        `\t* ${JSON.stringify(fieldName)}: ${JSON.stringify(fieldValue)}`
      );
    }
  }
}

const printTableInfo = (table, text) => {
  // Print header row
  let headerRowText = '';
  for (const headerCell of table.headerRows[0].cells) {
    const headerCellText = getText(headerCell.layout.textAnchor, text);
    headerRowText += `${JSON.stringify(headerCellText.trim())} | `;
  }
  console.log(
    `Collumns: ${headerRowText.substring(0, headerRowText.length - 3)}`
  );
  // Print first body row
  let bodyRowText = '';
  for (const bodyCell of table.bodyRows[0].cells) {
    const bodyCellText = getText(bodyCell.layout.textAnchor, text);
    bodyRowText += `${JSON.stringify(bodyCellText.trim())} | `;
  }
  console.log(
    `First row data: ${bodyRowText.substring(0, bodyRowText.length - 3)}`
  );
};

// Extract shards from the text field
const getText = (textAnchor, text) => {
  if (!textAnchor.textSegments || textAnchor.textSegments.length === 0) {
    return '';
  }

  // First shard in document doesn't have startIndex property
  const startIndex = textAnchor.textSegments[0].startIndex || 0;
  const endIndex = textAnchor.textSegments[0].endIndex;

  return text.substring(startIndex, endIndex);
};

Python

如需了解详情，请参阅 Document AI Python API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

from typing import Optional, Sequence

from google.api_core.client_options import ClientOptions
from google.cloud import documentai

# TODO(developer): Uncomment these variables before running the sample.
# project_id = "YOUR_PROJECT_ID"
# location = "YOUR_PROCESSOR_LOCATION" # Format is "us" or "eu"
# processor_id = "YOUR_PROCESSOR_ID" # Create processor before running sample
# processor_version = "rc" # Refer to https://cloud.google.com/document-ai/docs/manage-processor-versions for more information
# file_path = "/path/to/local/pdf"
# mime_type = "application/pdf" # Refer to https://cloud.google.com/document-ai/docs/file-types for supported file types


def process_document_form_sample(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
) -> documentai.Document:
    # Online processing request to Document AI
    document = process_document(
        project_id, location, processor_id, processor_version, file_path, mime_type
    )

    # Read the table and form fields output from the processor
    # The form processor also contains OCR data. For more information
    # on how to parse OCR data please see the OCR sample.

    text = document.text
    print(f"Full document text: {repr(text)}\n")
    print(f"There are {len(document.pages)} page(s) in this document.")

    # Read the form fields and tables output from the processor
    for page in document.pages:
        print(f"\n\n**** Page {page.page_number} ****")

        print(f"\nFound {len(page.tables)} table(s):")
        for table in page.tables:
            num_columns = len(table.header_rows[0].cells)
            num_rows = len(table.body_rows)
            print(f"Table with {num_columns} columns and {num_rows} rows:")

            # Print header rows
            print("Columns:")
            print_table_rows(table.header_rows, text)
            # Print body rows
            print("Table body data:")
            print_table_rows(table.body_rows, text)

        print(f"\nFound {len(page.form_fields)} form field(s):")
        for field in page.form_fields:
            name = layout_to_text(field.field_name, text)
            value = layout_to_text(field.field_value, text)
            print(f"    * {repr(name.strip())}: {repr(value.strip())}")

    # Supported in version `pretrained-form-parser-v2.0-2022-11-10` and later.
    # For more information: https://cloud.google.com/document-ai/docs/form-parser
    if document.entities:
        print(f"Found {len(document.entities)} generic entities:")
        for entity in document.entities:
            print_entity(entity)
            # Print Nested Entities
            for prop in entity.properties:
                print_entity(prop)

    return document


def print_table_rows(
    table_rows: Sequence[documentai.Document.Page.Table.TableRow], text: str
) -> None:
    for table_row in table_rows:
        row_text = ""
        for cell in table_row.cells:
            cell_text = layout_to_text(cell.layout, text)
            row_text += f"{repr(cell_text.strip())} | "
        print(row_text)




def print_entity(entity: documentai.Document.Entity) -> None:
    # Fields detected. For a full list of fields for each processor see
    # the processor documentation:
    # https://cloud.google.com/document-ai/docs/processors-list
    key = entity.type_

    # Some other value formats in addition to text are available
    # e.g. dates: `entity.normalized_value.date_value.year`
    text_value = entity.text_anchor.content or entity.mention_text
    confidence = entity.confidence
    normalized_value = entity.normalized_value.text
    print(f"    * {repr(key)}: {repr(text_value)} ({confidence:.1%} confident)")

    if normalized_value:
        print(f"    * Normalized Value: {repr(normalized_value)}")




def process_document(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
    process_options: Optional[documentai.ProcessOptions] = None,
) -> documentai.Document:
    # You must set the `api_endpoint` if you use a location other than "us".
    client = documentai.DocumentProcessorServiceClient(
        client_options=ClientOptions(
            api_endpoint=f"{location}-documentai.googleapis.com"
        )
    )

    # The full resource name of the processor version, e.g.:
    # `projects/{project_id}/locations/{location}/processors/{processor_id}/processorVersions/{processor_version_id}`
    # You must create a processor before running this sample.
    name = client.processor_version_path(
        project_id, location, processor_id, processor_version
    )

    # Read the file into memory
    with open(file_path, "rb") as image:
        image_content = image.read()

    # Configure the process request
    request = documentai.ProcessRequest(
        name=name,
        raw_document=documentai.RawDocument(content=image_content, mime_type=mime_type),
        # Only supported for Document OCR processor
        process_options=process_options,
    )

    result = client.process_document(request=request)

    # For a full list of `Document` object attributes, reference this page:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1/Document
    return result.document




def layout_to_text(layout: documentai.Document.Page.Layout, text: str) -> str:
    """
    Document AI identifies text in different parts of the document by their
    offsets in the entirety of the document"s text. This function converts
    offsets to a string.
    """
    # If a text segment spans several lines, it will
    # be stored in different text segments.
    return "".join(
        text[int(segment.start_index) : int(segment.end_index)]
        for segment in layout.text_anchor.text_segments
    )

实体、嵌套实体和归一化值

许多专用处理器会提取基于明确定义的架构的结构化数据。例如，账单解析器会检测特定字段，例如 invoice_date 和 supplier_name。以下是账单示例：

以下是账单解析器返回的完整文档对象：

下载 JSON

以下是文档对象的一些重要部分：

• 检测到的字段：Entities 包含处理器能够检测到的字段，例如 invoice_date：

  {
   "entities": [
      {
        "textAnchor": {
          "textSegments": [
            {
              "startIndex": "14",
              "endIndex": "24"
            }
          ],
          "content": "2020/01/01"
        },
        "type": "invoice_date",
        "confidence": 0.9938466,
        "pageAnchor": { ... },
        "id": "2",
        "normalizedValue": {
          "text": "2020-01-01",
          "dateValue": {
            "year": 2020,
            "month": 1,
            "day": 1
          }
        }
      }
    ]
  }
  

  对于某些字段，处理器还会标准化值。在此示例中，日期已从 2020/01/01 归一化为 2020-01-01。

• 标准化：对于许多特定的受支持字段，处理器还会标准化值，并返回 entity。normalizedValue 字段会添加到通过每个实体的 textAnchor 获取的原始提取字段中。因此，它会规范化字面量文本，通常会将文本值拆分为子字段。例如，2024 年 9 月 1 日这样的日期将表示为：

  normalizedValue": {
    "text": "2020-09-01",
    "dateValue": {
      "year": 2024,
      "month": 9,
      "day": 1
  }

在此示例中，日期已从 2020/01/01 规范化为 2020-01-01，这是一种标准化格式，可减少后期处理并实现转换为所选格式。

注意：日期和币种会进行标准化处理。它们依赖于对文档位置信息的解析。例如，美国使用 mm/dd/yyyy，而不是 dd/mm/yyyy。

地址通常也会进行标准化，即将地址的各个元素拆分为单独的字段。数字通过将整数或浮点数作为 normalizedValue 进行标准化。

• 丰富：某些处理器和字段还支持丰富。例如，文档 Google Singapore 中的原始 supplier_name 已根据 Enterprise Knowledge Graph 归一化为 Google Asia Pacific, Singapore。另请注意，由于 Enterprise Knowledge Graph 包含有关 Google 的信息，因此 Document AI 会推断 supplier_address，即使示例文档中没有该信息也是如此。

注意：Enterprise Knowledge Graph 用于信息丰富化，力求准确，但无法保证完美无缺。

  {
    "entities": [
      {
        "textAnchor": {
          "textSegments": [ ... ],
          "content": "Google Singapore"
        },
        "type": "supplier_name",
        "confidence": 0.39170802,
        "pageAnchor": { ... },
        "id": "12",
        "normalizedValue": {
          "text": "Google Asia Pacific, Singapore"
        }
      },
      {
        "type": "supplier_address",
        "id": "17",
        "normalizedValue": {
          "text": "70 Pasir Panjang Rd #03-71 Mapletree Business City II Singapore 117371",
          "addressValue": {
            "regionCode": "SG",
            "languageCode": "en-US",
            "postalCode": "117371",
            "addressLines": [
              "70 Pasir Panjang Rd",
              "#03-71 Mapletree Business City II"
            ]
          }
        }
      }
    ]
  }

• 嵌套字段：您可以先将一个实体声明为父级，然后在该父级下创建子级实体，从而创建嵌套架构（字段）。父字段的解析响应包含父字段的 properties 元素中的子字段。在以下示例中，line_item 是一个父字段，它有两个子字段：line_item/description 和 line_item/quantity。

  {
    "entities": [
      {
        "textAnchor": { ... },
        "type": "line_item",
        "confidence": 1.0,
        "pageAnchor": { ... },
        "id": "19",
        "properties": [
          {
            "textAnchor": {
              "textSegments": [ ... ],
              "content": "Tool A"
            },
            "type": "line_item/description",
            "confidence": 0.3461604,
            "pageAnchor": { ... },
            "id": "20"
          },
          {
            "textAnchor": {
              "textSegments": [ ... ],
              "content": "500"
            },
            "type": "line_item/quantity",
            "confidence": 0.8077843,
            "pageAnchor": { ... },
            "id": "21",
            "normalizedValue": {
              "text": "500"
            }
          }
        ]
      }
    ]
  }
  

注意：子字段可能不包含其父字段的名称。按照惯例，某些预训练处理器会为子字段添加 {parent}/ 前缀，例如 line_item/description 和 line_item/quantity，但并非所有处理器都遵循该惯例。

以下解析器会遵循该规范：

• 提取（自定义提取器）
• 旧版
  • 银行对账单解析器
  • 费用解析器
  • 账单解析器
  • 工资单解析器
  • W2 解析器

代码示例

以下代码示例演示了如何发送处理请求，然后将专用处理器的字段读取并输出到终端：

Java

如需了解详情，请参阅 Document AI Java API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

import com.google.cloud.documentai.v1.Document;
import com.google.cloud.documentai.v1.DocumentProcessorServiceClient;
import com.google.cloud.documentai.v1.DocumentProcessorServiceSettings;
import com.google.cloud.documentai.v1.ProcessRequest;
import com.google.cloud.documentai.v1.ProcessResponse;
import com.google.cloud.documentai.v1.RawDocument;
import com.google.protobuf.ByteString;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.TimeoutException;

public class ProcessSpecializedDocument {
  public static void processSpecializedDocument()
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // TODO(developer): Replace these variables before running the sample.
    String projectId = "your-project-id";
    String location = "your-project-location"; // Format is "us" or "eu".
    String processerId = "your-processor-id";
    String filePath = "path/to/input/file.pdf";
    processSpecializedDocument(projectId, location, processerId, filePath);
  }

  public static void processSpecializedDocument(
      String projectId, String location, String processorId, String filePath)
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // Initialize client that will be used to send requests. This client only needs
    // to be created
    // once, and can be reused for multiple requests. After completing all of your
    // requests, call
    // the "close" method on the client to safely clean up any remaining background
    // resources.
    String endpoint = String.format("%s-documentai.googleapis.com:443", location);
    DocumentProcessorServiceSettings settings =
        DocumentProcessorServiceSettings.newBuilder().setEndpoint(endpoint).build();
    try (DocumentProcessorServiceClient client = DocumentProcessorServiceClient.create(settings)) {
      // The full resource name of the processor, e.g.:
      // projects/project-id/locations/location/processor/processor-id
      // You must create new processors in the Cloud Console first
      String name =
          String.format("projects/%s/locations/%s/processors/%s", projectId, location, processorId);

      // Read the file.
      byte[] imageFileData = Files.readAllBytes(Paths.get(filePath));

      // Convert the image data to a Buffer and base64 encode it.
      ByteString content = ByteString.copyFrom(imageFileData);

      RawDocument document =
          RawDocument.newBuilder().setContent(content).setMimeType("application/pdf").build();

      // Configure the process request.
      ProcessRequest request =
          ProcessRequest.newBuilder().setName(name).setRawDocument(document).build();

      // Recognizes text entities in the PDF document
      ProcessResponse result = client.processDocument(request);
      Document documentResponse = result.getDocument();

      System.out.println("Document processing complete.");

      // Read fields specifically from the specialized US drivers license processor:
      // https://cloud.google.com/document-ai/docs/processors-list#processor_us-driver-license-parser
      // retrieving data from other specialized processors follow a similar pattern.
      // For a complete list of processors see:
      // https://cloud.google.com/document-ai/docs/processors-list
      //
      // OCR and other data is also present in the quality processor's response.
      // Please see the OCR and other samples for how to parse other data in the
      // response.
      for (Document.Entity entity : documentResponse.getEntitiesList()) {
        // Fields detected. For a full list of fields for each processor see
        // the processor documentation:
        // https://cloud.google.com/document-ai/docs/processors-list
        String entityType = entity.getType();
        // some other value formats in addition to text are available
        // e.g. dates: `entity.getNormalizedValue().getDateValue().getYear()`
        // check for normalized value with `entity.hasNormalizedValue()`
        String entityTextValue = escapeNewlines(entity.getTextAnchor().getContent());
        float entityConfidence = entity.getConfidence();
        System.out.printf(
            "    * %s: %s (%.2f%% confident)\n",
            entityType, entityTextValue, entityConfidence * 100.0);
      }
    }
  }

  private static String escapeNewlines(String s) {
    return s.replace("\n", "\\n").replace("\r", "\\r");
  }
}

Node.js

如需了解详情，请参阅 Document AI Node.js API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

/**
 * TODO(developer): Uncomment these variables before running the sample.
 */
// const projectId = 'YOUR_PROJECT_ID';
// const location = 'YOUR_PROJECT_LOCATION'; // Format is 'us' or 'eu'
// const processorId = 'YOUR_PROCESSOR_ID'; // Create processor in Cloud Console
// const filePath = '/path/to/local/pdf';

const {DocumentProcessorServiceClient} =
  require('@google-cloud/documentai').v1beta3;

// Instantiates a client
const client = new DocumentProcessorServiceClient();

async function processDocument() {
  // The full resource name of the processor, e.g.:
  // projects/project-id/locations/location/processor/processor-id
  // You must create new processors in the Cloud Console first
  const name = `projects/${projectId}/locations/${location}/processors/${processorId}`;

  // Read the file into memory.
  const fs = require('fs').promises;
  const imageFile = await fs.readFile(filePath);

  // Convert the image data to a Buffer and base64 encode it.
  const encodedImage = Buffer.from(imageFile).toString('base64');

  const request = {
    name,
    rawDocument: {
      content: encodedImage,
      mimeType: 'application/pdf',
    },
  };

  // Recognizes text entities in the PDF document
  const [result] = await client.processDocument(request);

  console.log('Document processing complete.');

  // Read fields specificly from the specalized US drivers license processor:
  // https://cloud.google.com/document-ai/docs/processors-list#processor_us-driver-license-parser
  // retriving data from other specalized processors follow a similar pattern.
  // For a complete list of processors see:
  // https://cloud.google.com/document-ai/docs/processors-list
  //
  // OCR and other data is also present in the quality processor's response.
  // Please see the OCR and other samples for how to parse other data in the
  // response.
  const {document} = result;
  for (const entity of document.entities) {
    // Fields detected. For a full list of fields for each processor see
    // the processor documentation:
    // https://cloud.google.com/document-ai/docs/processors-list
    const key = entity.type;
    // some other value formats in addition to text are availible
    // e.g. dates: `entity.normalizedValue.dateValue.year`
    const textValue =
      entity.textAnchor !== null ? entity.textAnchor.content : '';
    const conf = entity.confidence * 100;
    console.log(
      `* ${JSON.stringify(key)}: ${JSON.stringify(textValue)}(${conf.toFixed(
        2
      )}% confident)`
    );
  }
}

Python

如需了解详情，请参阅 Document AI Python API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

from typing import Optional, Sequence

from google.api_core.client_options import ClientOptions
from google.cloud import documentai

# TODO(developer): Uncomment these variables before running the sample.
# project_id = "YOUR_PROJECT_ID"
# location = "YOUR_PROCESSOR_LOCATION" # Format is "us" or "eu"
# processor_id = "YOUR_PROCESSOR_ID" # Create processor before running sample
# processor_version = "rc" # Refer to https://cloud.google.com/document-ai/docs/manage-processor-versions for more information
# file_path = "/path/to/local/pdf"
# mime_type = "application/pdf" # Refer to https://cloud.google.com/document-ai/docs/file-types for supported file types


def process_document_entity_extraction_sample(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
) -> None:
    # Online processing request to Document AI
    document = process_document(
        project_id, location, processor_id, processor_version, file_path, mime_type
    )

    # Print extracted entities from entity extraction processor output.
    # For a complete list of processors see:
    # https://cloud.google.com/document-ai/docs/processors-list
    #
    # OCR and other data is also present in the processor's response.
    # Refer to the OCR samples for how to parse other data in the response.

    print(f"Found {len(document.entities)} entities:")
    for entity in document.entities:
        print_entity(entity)
        # Print Nested Entities (if any)
        for prop in entity.properties:
            print_entity(prop)




def print_entity(entity: documentai.Document.Entity) -> None:
    # Fields detected. For a full list of fields for each processor see
    # the processor documentation:
    # https://cloud.google.com/document-ai/docs/processors-list
    key = entity.type_

    # Some other value formats in addition to text are available
    # e.g. dates: `entity.normalized_value.date_value.year`
    text_value = entity.text_anchor.content or entity.mention_text
    confidence = entity.confidence
    normalized_value = entity.normalized_value.text
    print(f"    * {repr(key)}: {repr(text_value)} ({confidence:.1%} confident)")

    if normalized_value:
        print(f"    * Normalized Value: {repr(normalized_value)}")




def process_document(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
    process_options: Optional[documentai.ProcessOptions] = None,
) -> documentai.Document:
    # You must set the `api_endpoint` if you use a location other than "us".
    client = documentai.DocumentProcessorServiceClient(
        client_options=ClientOptions(
            api_endpoint=f"{location}-documentai.googleapis.com"
        )
    )

    # The full resource name of the processor version, e.g.:
    # `projects/{project_id}/locations/{location}/processors/{processor_id}/processorVersions/{processor_version_id}`
    # You must create a processor before running this sample.
    name = client.processor_version_path(
        project_id, location, processor_id, processor_version
    )

    # Read the file into memory
    with open(file_path, "rb") as image:
        image_content = image.read()

    # Configure the process request
    request = documentai.ProcessRequest(
        name=name,
        raw_document=documentai.RawDocument(content=image_content, mime_type=mime_type),
        # Only supported for Document OCR processor
        process_options=process_options,
    )

    result = client.process_document(request=request)

    # For a full list of `Document` object attributes, reference this page:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1/Document
    return result.document

自定义文档提取器

自定义文档提取器处理器可以从没有预训练处理器的文档中提取自定义实体。这可以通过训练自定义模型或使用生成式 AI 基础模型来提取命名实体（无需任何训练）来实现。如需了解详情，请参阅在控制台中创建自定义文档提取器。

• 如果您训练自定义模型，则该处理器的使用方式与预训练的实体提取处理器完全相同。
• 如果您使用基础模型，则可以创建处理器版本，以便为每个请求提取特定实体，也可以按请求进行配置。

如需了解输出结构，请参阅实体、嵌套实体和标准化值。

代码示例

如果您使用的是自定义模型，或者使用基础模型创建了处理器版本，请使用实体提取代码示例。

以下代码示例演示了如何根据每个请求为基础模型自定义文档提取器配置特定实体，并输出提取的实体：

Python

如需了解详情，请参阅 Document AI Python API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

from typing import Optional, Sequence

from google.api_core.client_options import ClientOptions
from google.cloud import documentai

# TODO(developer): Uncomment these variables before running the sample.
# project_id = "YOUR_PROJECT_ID"
# location = "YOUR_PROCESSOR_LOCATION" # Format is "us" or "eu"
# processor_id = "YOUR_PROCESSOR_ID" # Create processor before running sample
# processor_version = "rc" # Refer to https://cloud.google.com/document-ai/docs/manage-processor-versions for more information
# file_path = "/path/to/local/pdf"
# mime_type = "application/pdf" # Refer to https://cloud.google.com/document-ai/docs/file-types for supported file types




def process_document_custom_extractor_sample(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
) -> None:
    # Entities to extract from Foundation Model CDE
    properties = [
        documentai.DocumentSchema.EntityType.Property(
            name="invoice_id",
            value_type="string",
            occurrence_type=documentai.DocumentSchema.EntityType.Property.OccurrenceType.REQUIRED_ONCE,
        ),
        documentai.DocumentSchema.EntityType.Property(
            name="notes",
            value_type="string",
            occurrence_type=documentai.DocumentSchema.EntityType.Property.OccurrenceType.OPTIONAL_MULTIPLE,
        ),
        documentai.DocumentSchema.EntityType.Property(
            name="terms",
            value_type="string",
            occurrence_type=documentai.DocumentSchema.EntityType.Property.OccurrenceType.OPTIONAL_MULTIPLE,
        ),
    ]
    # Optional: For Generative AI processors, request different fields than the
    # schema for a processor version
    process_options = documentai.ProcessOptions(
        schema_override=documentai.DocumentSchema(
            display_name="CDE Schema",
            description="Document Schema for the CDE Processor",
            entity_types=[
                documentai.DocumentSchema.EntityType(
                    name="custom_extraction_document_type",
                    base_types=["document"],
                    properties=properties,
                )
            ],
        )
    )

    # Online processing request to Document AI
    document = process_document(
        project_id,
        location,
        processor_id,
        processor_version,
        file_path,
        mime_type,
        process_options=process_options,
    )

    for entity in document.entities:
        print_entity(entity)
        # Print Nested Entities (if any)
        for prop in entity.properties:
            print_entity(prop)




def print_entity(entity: documentai.Document.Entity) -> None:
    # Fields detected. For a full list of fields for each processor see
    # the processor documentation:
    # https://cloud.google.com/document-ai/docs/processors-list
    key = entity.type_

    # Some other value formats in addition to text are available
    # e.g. dates: `entity.normalized_value.date_value.year`
    text_value = entity.text_anchor.content or entity.mention_text
    confidence = entity.confidence
    normalized_value = entity.normalized_value.text
    print(f"    * {repr(key)}: {repr(text_value)} ({confidence:.1%} confident)")

    if normalized_value:
        print(f"    * Normalized Value: {repr(normalized_value)}")




def process_document(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
    process_options: Optional[documentai.ProcessOptions] = None,
) -> documentai.Document:
    # You must set the `api_endpoint` if you use a location other than "us".
    client = documentai.DocumentProcessorServiceClient(
        client_options=ClientOptions(
            api_endpoint=f"{location}-documentai.googleapis.com"
        )
    )

    # The full resource name of the processor version, e.g.:
    # `projects/{project_id}/locations/{location}/processors/{processor_id}/processorVersions/{processor_version_id}`
    # You must create a processor before running this sample.
    name = client.processor_version_path(
        project_id, location, processor_id, processor_version
    )

    # Read the file into memory
    with open(file_path, "rb") as image:
        image_content = image.read()

    # Configure the process request
    request = documentai.ProcessRequest(
        name=name,
        raw_document=documentai.RawDocument(content=image_content, mime_type=mime_type),
        # Only supported for Document OCR processor
        process_options=process_options,
    )

    result = client.process_document(request=request)

    # For a full list of `Document` object attributes, reference this page:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1/Document
    return result.document

摘要

摘要器处理器使用生成式 AI 基础模型来汇总从文档中提取的文本。您可以通过以下方式自定义回答的长度和格式：

• 长度
  • BRIEF：一两句话的简短摘要
  • MODERATE：段落长度的摘要
  • COMPREHENSIVE：可用的最长选项
• 格式
  • PARAGRAPH：以段落形式输出
  • BULLETS：以项目符号列表的形式输出

您可以为特定长度和格式创建处理器版本，也可以按请求进行配置。

摘要文本会显示在 Document.entities.normalizedValue.text 中。您可以在处理器输出示例中找到完整的示例输出 JSON 文件。

如需了解详情，请参阅在控制台中构建文档摘要生成器。

代码示例

以下代码示例演示了如何在处理请求中配置特定长度和格式，并输出摘要文本：

Python

如需了解详情，请参阅 Document AI Python API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

from typing import Optional

from google.api_core.client_options import ClientOptions
from google.cloud import documentai_v1beta3 as documentai


# TODO(developer): Uncomment these variables before running the sample.
# project_id = "YOUR_PROJECT_ID"
# location = "YOUR_PROCESSOR_LOCATION" # Format is "us" or "eu"
# processor_id = "YOUR_PROCESSOR_ID" # Create processor before running sample
# processor_version = "rc" # Refer to https://cloud.google.com/document-ai/docs/manage-processor-versions for more information
# file_path = "/path/to/local/pdf"
# mime_type = "application/pdf" # Refer to https://cloud.google.com/document-ai/docs/file-types for supported file types

def process_document_summarizer_sample(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
) -> None:
    # For supported options, refer to:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1beta3/projects.locations.processors.processorVersions#summaryoptions
    summary_options = documentai.SummaryOptions(
        length=documentai.SummaryOptions.Length.BRIEF,
        format=documentai.SummaryOptions.Format.BULLETS,
    )

    properties = [
        documentai.DocumentSchema.EntityType.Property(
            name="summary",
            value_type="string",
            occurrence_type=documentai.DocumentSchema.EntityType.Property.OccurrenceType.REQUIRED_ONCE,
            property_metadata=documentai.PropertyMetadata(
                field_extraction_metadata=documentai.FieldExtractionMetadata(
                    summary_options=summary_options
                )
            ),
        )
    ]

    # Optional: Request specific summarization format other than the default
    # for the processor version.
    process_options = documentai.ProcessOptions(
        schema_override=documentai.DocumentSchema(
            entity_types=[
                documentai.DocumentSchema.EntityType(
                    name="summary_document_type",
                    base_types=["document"],
                    properties=properties,
                )
            ]
        )
    )

    # Online processing request to Document AI
    document = process_document(
        project_id,
        location,
        processor_id,
        processor_version,
        file_path,
        mime_type,
        process_options=process_options,
    )

    for entity in document.entities:
        print_entity(entity)
        # Print Nested Entities (if any)
        for prop in entity.properties:
            print_entity(prop)


def print_entity(entity: documentai.Document.Entity) -> None:
    # Fields detected. For a full list of fields for each processor see
    # the processor documentation:
    # https://cloud.google.com/document-ai/docs/processors-list
    key = entity.type_

    # Some other value formats in addition to text are availible
    # e.g. dates: `entity.normalized_value.date_value.year`
    text_value = entity.text_anchor.content
    confidence = entity.confidence
    normalized_value = entity.normalized_value.text
    print(f"    * {repr(key)}: {repr(text_value)}({confidence:.1%} confident)")

    if normalized_value:
        print(f"    * Normalized Value: {repr(normalized_value)}")


def process_document(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
    process_options: Optional[documentai.ProcessOptions] = None,
) -> documentai.Document:
    # You must set the `api_endpoint` if you use a location other than "us".
    client = documentai.DocumentProcessorServiceClient(
        client_options=ClientOptions(
            api_endpoint=f"{location}-documentai.googleapis.com"
        )
    )

    # The full resource name of the processor version, e.g.:
    # `projects/{project_id}/locations/{location}/processors/{processor_id}/processorVersions/{processor_version_id}`
    # You must create a processor before running this sample.
    name = client.processor_version_path(
        project_id, location, processor_id, processor_version
    )

    # Read the file into memory
    with open(file_path, "rb") as image:
        image_content = image.read()

    # Configure the process request
    request = documentai.ProcessRequest(
        name=name,
        raw_document=documentai.RawDocument(content=image_content, mime_type=mime_type),
        # Only supported for Document OCR processor
        process_options=process_options,
    )

    result = client.process_document(request=request)

    # For a full list of `Document` object attributes, reference this page:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1/Document
    return result.document

拆分和分类

以下是一个包含不同类型文档和表单的 10 页复合 PDF 文件：

下载 PDF 文件

以下是借贷单据拆分器和分类器返回的完整文档对象：

下载 JSON

拆分器检测到的每个文档都由一个 entity 表示。例如：

  {
    "entities": [
      {
        "textAnchor": {
          "textSegments": [
            {
              "startIndex": "13936",
              "endIndex": "21108"
            }
          ]
        },
        "type": "1040se_2020",
        "confidence": 0.76257163,
        "pageAnchor": {
          "pageRefs": [
            {
              "page": "6"
            },
            {
              "page": "7"
            }
          ]
        }
      }
    ]
  }

• Entity.pageAnchor 表示此文档有 2 页。请注意，pageRefs[].page 从零开始，是 document.pages[] 字段的索引。

  注意：零 pageRefs[].page 值会被省略。当 API 检测到页面值为 "0" 时，该坐标值将在 JSON 响应中被省略。例如，文档前两页的响应将为："pageRefs": [{}, {"page": "1"}]

• Entity.type 指定此文档是 1040 Schedule SE 表单。如需查看可识别的文档类型的完整列表，请参阅处理器文档中的识别的文档类型。

如需了解详情，请参阅文档拆分器行为。

代码示例

拆分器会识别页面边界，但不会实际为您拆分输入文档。您可以使用 Document AI 工具箱，通过页面边界对 PDF 文件进行物理拆分。 以下代码示例会输出页面范围，而不会拆分 PDF：

Java

如需了解详情，请参阅 Document AI Java API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

import com.google.cloud.documentai.v1.Document;
import com.google.cloud.documentai.v1.DocumentProcessorServiceClient;
import com.google.cloud.documentai.v1.DocumentProcessorServiceSettings;
import com.google.cloud.documentai.v1.ProcessRequest;
import com.google.cloud.documentai.v1.ProcessResponse;
import com.google.cloud.documentai.v1.RawDocument;
import com.google.protobuf.ByteString;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
import java.util.List;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.TimeoutException;

public class ProcessSplitterDocument {
  public static void processSplitterDocument()
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // TODO(developer): Replace these variables before running the sample.
    String projectId = "your-project-id";
    String location = "your-project-location"; // Format is "us" or "eu".
    String processerId = "your-processor-id";
    String filePath = "path/to/input/file.pdf";
    processSplitterDocument(projectId, location, processerId, filePath);
  }

  public static void processSplitterDocument(
      String projectId, String location, String processorId, String filePath)
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // Initialize client that will be used to send requests. This client only needs
    // to be created
    // once, and can be reused for multiple requests. After completing all of your
    // requests, call
    // the "close" method on the client to safely clean up any remaining background
    // resources.
    String endpoint = String.format("%s-documentai.googleapis.com:443", location);
    DocumentProcessorServiceSettings settings =
        DocumentProcessorServiceSettings.newBuilder().setEndpoint(endpoint).build();
    try (DocumentProcessorServiceClient client = DocumentProcessorServiceClient.create(settings)) {
      // The full resource name of the processor, e.g.:
      // projects/project-id/locations/location/processor/processor-id
      // You must create new processors in the Cloud Console first
      String name =
          String.format("projects/%s/locations/%s/processors/%s", projectId, location, processorId);

      // Read the file.
      byte[] imageFileData = Files.readAllBytes(Paths.get(filePath));

      // Convert the image data to a Buffer and base64 encode it.
      ByteString content = ByteString.copyFrom(imageFileData);

      RawDocument document =
          RawDocument.newBuilder().setContent(content).setMimeType("application/pdf").build();

      // Configure the process request.
      ProcessRequest request =
          ProcessRequest.newBuilder().setName(name).setRawDocument(document).build();

      // Recognizes text entities in the PDF document
      ProcessResponse result = client.processDocument(request);
      Document documentResponse = result.getDocument();

      System.out.println("Document processing complete.");

      // Read the splitter output from the document splitter processor:
      // https://cloud.google.com/document-ai/docs/processors-list#processor_doc-splitter
      // This processor only provides text for the document and information on how
      // to split the document on logical boundaries. To identify and extract text,
      // form elements, and entities please see other processors like the OCR, form,
      // and specialized processors.
      List<Document.Entity> entities = documentResponse.getEntitiesList();
      System.out.printf("Found %d subdocuments:\n", entities.size());
      for (Document.Entity entity : entities) {
        float entityConfidence = entity.getConfidence();
        String pagesRangeText = pageRefsToString(entity.getPageAnchor().getPageRefsList());
        String subdocumentType = entity.getType();
        if (subdocumentType.isEmpty()) {
          System.out.printf(
              "%.2f%% confident that %s a subdocument.\n", entityConfidence * 100, pagesRangeText);
        } else {
          System.out.printf(
              "%.2f%% confident that %s a '%s' subdocument.\n",
              entityConfidence * 100, pagesRangeText, subdocumentType);
        }
      }
    }
  }

  // Converts page reference(s) to a string describing the page or page range.
  private static String pageRefsToString(List<Document.PageAnchor.PageRef> pageRefs) {
    if (pageRefs.size() == 1) {
      return String.format("page %d is", pageRefs.get(0).getPage() + 1);
    } else {
      long start = pageRefs.get(0).getPage() + 1;
      long end = pageRefs.get(1).getPage() + 1;
      return String.format("pages %d to %d are", start, end);
    }
  }
}

Node.js

如需了解详情，请参阅 Document AI Node.js API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

/**
 * TODO(developer): Uncomment these variables before running the sample.
 */
// const projectId = 'YOUR_PROJECT_ID';
// const location = 'YOUR_PROJECT_LOCATION'; // Format is 'us' or 'eu'
// const processorId = 'YOUR_PROCESSOR_ID'; // Create processor in Cloud Console
// const filePath = '/path/to/local/pdf';

const {DocumentProcessorServiceClient} =
  require('@google-cloud/documentai').v1beta3;

// Instantiates a client
const client = new DocumentProcessorServiceClient();

async function processDocument() {
  // The full resource name of the processor, e.g.:
  // projects/project-id/locations/location/processor/processor-id
  // You must create new processors in the Cloud Console first
  const name = `projects/${projectId}/locations/${location}/processors/${processorId}`;

  // Read the file into memory.
  const fs = require('fs').promises;
  const imageFile = await fs.readFile(filePath);

  // Convert the image data to a Buffer and base64 encode it.
  const encodedImage = Buffer.from(imageFile).toString('base64');

  const request = {
    name,
    rawDocument: {
      content: encodedImage,
      mimeType: 'application/pdf',
    },
  };

  // Recognizes text entities in the PDF document
  const [result] = await client.processDocument(request);

  console.log('Document processing complete.');

  // Read fields specificly from the specalized US drivers license processor:
  // https://cloud.google.com/document-ai/docs/processors-list#processor_us-driver-license-parser
  // retriving data from other specalized processors follow a similar pattern.
  // For a complete list of processors see:
  // https://cloud.google.com/document-ai/docs/processors-list
  //
  // OCR and other data is also present in the quality processor's response.
  // Please see the OCR and other samples for how to parse other data in the
  // response.
  const {document} = result;
  console.log(`Found ${document.entities.length} subdocuments:`);
  for (const entity of document.entities) {
    const conf = entity.confidence * 100;
    const pagesRange = pageRefsToRange(entity.pageAnchor.pageRefs);
    if (entity.type !== '') {
      console.log(
        `${conf.toFixed(2)}% confident that ${pagesRange} a "${
          entity.type
        }" subdocument.`
      );
    } else {
      console.log(
        `${conf.toFixed(2)}% confident that ${pagesRange} a subdocument.`
      );
    }
  }
}

// Converts a page ref to a string describing the page or page range.
const pageRefsToRange = pageRefs => {
  if (pageRefs.length === 1) {
    const num = parseInt(pageRefs[0].page) + 1 || 1;
    return `page ${num} is`;
  } else {
    const start = parseInt(pageRefs[0].page) + 1 || 1;
    const end = parseInt(pageRefs[1].page) + 1;
    return `pages ${start} to ${end} are`;
  }
};

Python

如需了解详情，请参阅 Document AI Python API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

from typing import Optional, Sequence

from google.api_core.client_options import ClientOptions
from google.cloud import documentai

# TODO(developer): Uncomment these variables before running the sample.
# project_id = "YOUR_PROJECT_ID"
# location = "YOUR_PROCESSOR_LOCATION" # Format is "us" or "eu"
# processor_id = "YOUR_PROCESSOR_ID" # Create processor before running sample
# processor_version = "rc" # Refer to https://cloud.google.com/document-ai/docs/manage-processor-versions for more information
# file_path = "/path/to/local/pdf"
# mime_type = "application/pdf" # Refer to https://cloud.google.com/document-ai/docs/file-types for supported file types


def process_document_splitter_sample(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
) -> None:
    # Online processing request to Document AI
    document = process_document(
        project_id, location, processor_id, processor_version, file_path, mime_type
    )

    # Read the splitter output from a document splitter/classifier processor:
    # e.g. https://cloud.google.com/document-ai/docs/processors-list#processor_procurement-document-splitter
    # This processor only provides text for the document and information on how
    # to split the document on logical boundaries. To identify and extract text,
    # form elements, and entities please see other processors like the OCR, form,
    # and specalized processors.

    print(f"Found {len(document.entities)} subdocuments:")
    for entity in document.entities:
        conf_percent = f"{entity.confidence:.1%}"
        pages_range = page_refs_to_string(entity.page_anchor.page_refs)

        # Print subdocument type information, if available
        if entity.type_:
            print(
                f"{conf_percent} confident that {pages_range} a '{entity.type_}' subdocument."
            )
        else:
            print(f"{conf_percent} confident that {pages_range} a subdocument.")


def page_refs_to_string(
    page_refs: Sequence[documentai.Document.PageAnchor.PageRef],
) -> str:
    """Converts a page ref to a string describing the page or page range."""
    pages = [str(int(page_ref.page) + 1) for page_ref in page_refs]
    if len(pages) == 1:
        return f"page {pages[0]} is"
    else:
        return f"pages {', '.join(pages)} are"




def process_document(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
    process_options: Optional[documentai.ProcessOptions] = None,
) -> documentai.Document:
    # You must set the `api_endpoint` if you use a location other than "us".
    client = documentai.DocumentProcessorServiceClient(
        client_options=ClientOptions(
            api_endpoint=f"{location}-documentai.googleapis.com"
        )
    )

    # The full resource name of the processor version, e.g.:
    # `projects/{project_id}/locations/{location}/processors/{processor_id}/processorVersions/{processor_version_id}`
    # You must create a processor before running this sample.
    name = client.processor_version_path(
        project_id, location, processor_id, processor_version
    )

    # Read the file into memory
    with open(file_path, "rb") as image:
        image_content = image.read()

    # Configure the process request
    request = documentai.ProcessRequest(
        name=name,
        raw_document=documentai.RawDocument(content=image_content, mime_type=mime_type),
        # Only supported for Document OCR processor
        process_options=process_options,
    )

    result = client.process_document(request=request)

    # For a full list of `Document` object attributes, reference this page:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1/Document
    return result.document

以下代码示例使用 Document AI Toolbox，根据已处理的 Document 中的页面边界来拆分 PDF 文件。

Python

如需了解详情，请参阅 Document AI Python API 参考文档。

如需向 Document AI 进行身份验证，请设置应用默认凭据。 如需了解详情，请参阅为本地开发环境设置身份验证。

from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a local document.proto or sharded document.proto from a splitter/classifier in path
# document_path = "path/to/local/document.json"
# pdf_path = "path/to/local/document.pdf"
# output_path = "resources/output/"


def split_pdf_sample(document_path: str, pdf_path: str, output_path: str) -> None:
    wrapped_document = document.Document.from_document_path(document_path=document_path)

    output_files = wrapped_document.split_pdf(
        pdf_path=pdf_path, output_path=output_path
    )

    print("Document Successfully Split")
    for output_file in output_files:
        print(output_file)

Document AI Toolbox

实验性

此库受服务专用条款的“通用服务条款”部分中的“非正式版产品条款”的约束。 非正式版库“按原样”提供，可能只能获得有限的支持。 如需了解详情，请参阅发布阶段说明。

Document AI Toolbox 是一个 Python 版 SDK，提供用于管理、操作和从文档响应中提取信息的实用函数。它会根据 Cloud Storage 中的 JSON 文件、本地 JSON 文件或直接从 process_document() 方法输出的处理后文档响应创建“封装”文档对象。

它可以执行以下操作：

• 将批处理中分散的 Document JSON 文件合并为一个“封装”文档。
• 将分片导出为统一的 Document。
• 从以下位置获取 Document 输出：
  • Cloud Storage
  • BatchProcessMetadata
  • Operation 名称
• 无需处理 Layout 信息，即可访问 Pages、Lines、Paragraphs、FormFields 和 Tables 中的文本。
• 搜索包含目标字符串或与正则表达式匹配的 Pages。
• 按名称搜索 FormFields。
• 按类型搜索 Entities。
• 将 Tables 转换为 Pandas Dataframe 或 CSV。
• 将 Entities 和 FormFields 插入到 BigQuery 表中。
• 根据拆分器/分类器处理器的输出拆分 PDF 文件。
• 从 Document 边界框中提取图片 Entities。
• 在 Documents 与常用格式之间进行转换：
  • Cloud Vision API AnnotateFileResponse
  • hOCR
  • 第三方文档处理格式
  注意：导入文档并将其注释转换为 Document AI 格式时，这些文档在 Google Cloud 控制台中会被归类为 Labeled。
• 从 Cloud Storage 文件夹中创建要处理的文档批次。

代码示例

以下代码示例演示了如何使用 Document AI Toolbox。

快速入门

from typing import Optional

from google.cloud import documentai
from google.cloud.documentai_toolbox import document, gcs_utilities

# TODO(developer): Uncomment these variables before running the sample.
# Given a Document JSON or sharded Document JSON in path gs://bucket/path/to/folder
# gcs_bucket_name = "bucket"
# gcs_prefix = "path/to/folder"

# Or, given a Document JSON in path gs://bucket/path/to/folder/document.json
# gcs_uri = "gs://bucket/path/to/folder/document.json"

# Or, given a Document JSON in path local/path/to/folder/document.json
# document_path = "local/path/to/folder/document.json"

# Or, given a Document object from Document AI
# documentai_document = documentai.Document()

# Or, given a BatchProcessMetadata object from Document AI
# operation = client.batch_process_documents(request)
# operation.result(timeout=timeout)
# batch_process_metadata = documentai.BatchProcessMetadata(operation.metadata)

# Or, given a BatchProcessOperation name from Document AI
# batch_process_operation = "projects/project_id/locations/location/operations/operation_id"


def quickstart_sample(
    gcs_bucket_name: Optional[str] = None,
    gcs_prefix: Optional[str] = None,
    gcs_uri: Optional[str] = None,
    document_path: Optional[str] = None,
    documentai_document: Optional[documentai.Document] = None,
    batch_process_metadata: Optional[documentai.BatchProcessMetadata] = None,
    batch_process_operation: Optional[str] = None,
) -> document.Document:
    if gcs_bucket_name and gcs_prefix:
        # Load from Google Cloud Storage Directory
        print("Document structure in Cloud Storage")
        gcs_utilities.print_gcs_document_tree(
            gcs_bucket_name=gcs_bucket_name, gcs_prefix=gcs_prefix
        )

        wrapped_document = document.Document.from_gcs(
            gcs_bucket_name=gcs_bucket_name, gcs_prefix=gcs_prefix
        )
    elif gcs_uri:
        # Load a single Document from a Google Cloud Storage URI
        wrapped_document = document.Document.from_gcs_uri(gcs_uri=gcs_uri)
    elif document_path:
        # Load from local `Document` JSON file
        wrapped_document = document.Document.from_document_path(document_path)
    elif documentai_document:
        # Load from `documentai.Document` object
        wrapped_document = document.Document.from_documentai_document(
            documentai_document
        )
    elif batch_process_metadata:
        # Load Documents from `BatchProcessMetadata` object
        wrapped_documents = document.Document.from_batch_process_metadata(
            metadata=batch_process_metadata
        )
        wrapped_document = wrapped_documents[0]
    elif batch_process_operation:
        wrapped_documents = document.Document.from_batch_process_operation(
            location="us", operation_name=batch_process_operation
        )
        wrapped_document = wrapped_documents[0]
    else:
        raise ValueError("No document source provided.")

    # For all properties and methods, refer to:
    # https://cloud.google.com/python/docs/reference/documentai-toolbox/latest/google.cloud.documentai_toolbox.wrappers.document.Document

    print("Document Successfully Loaded!")
    print(f"\t Number of Pages: {len(wrapped_document.pages)}")
    print(f"\t Number of Entities: {len(wrapped_document.entities)}")

    for page in wrapped_document.pages:
        print(f"Page {page.page_number}")
        for block in page.blocks:
            print(block.text)
        for paragraph in page.paragraphs:
            print(paragraph.text)
        for line in page.lines:
            print(line.text)
        for token in page.tokens:
            print(token.text)

        # Only supported with Form Parser processor
        # https://cloud.google.com/document-ai/docs/form-parser
        for form_field in page.form_fields:
            print(f"{form_field.field_name} : {form_field.field_value}")

        # Only supported with Enterprise Document OCR version `pretrained-ocr-v2.0-2023-06-02`
        # https://cloud.google.com/document-ai/docs/process-documents-ocr#enable_symbols
        for symbol in page.symbols:
            print(symbol.text)

        # Only supported with Enterprise Document OCR version `pretrained-ocr-v2.0-2023-06-02`
        # https://cloud.google.com/document-ai/docs/process-documents-ocr#math_ocr
        for math_formula in page.math_formulas:
            print(math_formula.text)

    # Only supported with Entity Extraction processors
    # https://cloud.google.com/document-ai/docs/processors-list
    for entity in wrapped_document.entities:
        print(f"{entity.type_} : {entity.mention_text}")
        if entity.normalized_text:
            print(f"\tNormalized Text: {entity.normalized_text}")

    # Only supported with Layout Parser
    for chunk in wrapped_document.chunks:
        print(f"Chunk {chunk.chunk_id}: {chunk.content}")

    for block in wrapped_document.document_layout_blocks:
        print(f"Document Layout Block {block.block_id}")

        if block.text_block:
            print(f"{block.text_block.type_}: {block.text_block.text}")
        if block.list_block:
            print(f"{block.list_block.type_}: {block.list_block.list_entries}")
        if block.table_block:
            print(block.table_block.header_rows, block.table_block.body_rows)

表

from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a local document.proto or sharded document.proto in path
# document_path = "path/to/local/document.json"
# output_file_prefix = "output/table"


def table_sample(document_path: str, output_file_prefix: str) -> None:
    wrapped_document = document.Document.from_document_path(document_path=document_path)

    print("Tables in Document")
    for page in wrapped_document.pages:
        for table_index, table in enumerate(page.tables):
            # Convert table to Pandas Dataframe
            # Refer to https://pandas.pydata.org/docs/reference/frame.html for all supported methods
            df = table.to_dataframe()
            print(df)

            output_filename = f"{output_file_prefix}-{page.page_number}-{table_index}"

            # Write Dataframe to CSV file
            df.to_csv(f"{output_filename}.csv", index=False)

            # Write Dataframe to HTML file
            df.to_html(f"{output_filename}.html", index=False)

            # Write Dataframe to Markdown file
            df.to_markdown(f"{output_filename}.md", index=False)

BigQuery Export

from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a document.proto or sharded document.proto in path gs://bucket/path/to/folder
# gcs_bucket_name = "bucket"
# gcs_prefix = "path/to/folder"
# dataset_name = "test_dataset"
# table_name = "test_table"
# project_id = "YOUR_PROJECT_ID"


def entities_to_bigquery_sample(
    gcs_bucket_name: str,
    gcs_prefix: str,
    dataset_name: str,
    table_name: str,
    project_id: str,
) -> None:
    wrapped_document = document.Document.from_gcs(
        gcs_bucket_name=gcs_bucket_name, gcs_prefix=gcs_prefix
    )

    job = wrapped_document.entities_to_bigquery(
        dataset_name=dataset_name, table_name=table_name, project_id=project_id
    )

    # Also supported:
    # job = wrapped_document.form_fields_to_bigquery(
    #     dataset_name=dataset_name, table_name=table_name, project_id=project_id
    # )

    print("Document entities loaded into BigQuery")
    print(f"Job ID: {job.job_id}")
    print(f"Table: {job.destination.path}")

PDF 拆分

from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a local document.proto or sharded document.proto from a splitter/classifier in path
# document_path = "path/to/local/document.json"
# pdf_path = "path/to/local/document.pdf"
# output_path = "resources/output/"


def split_pdf_sample(document_path: str, pdf_path: str, output_path: str) -> None:
    wrapped_document = document.Document.from_document_path(document_path=document_path)

    output_files = wrapped_document.split_pdf(
        pdf_path=pdf_path, output_path=output_path
    )

    print("Document Successfully Split")
    for output_file in output_files:
        print(output_file)

图片提取

from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a local document.proto or sharded document.proto from an identity processor in path
# document_path = "path/to/local/document.json"
# output_path = "resources/output/"
# output_file_prefix = "exported_photo"
# output_file_extension = "png"


def export_images_sample(
    document_path: str,
    output_path: str,
    output_file_prefix: str,
    output_file_extension: str,
) -> None:
    wrapped_document = document.Document.from_document_path(document_path=document_path)

    output_files = wrapped_document.export_images(
        output_path=output_path,
        output_file_prefix=output_file_prefix,
        output_file_extension=output_file_extension,
    )
    print("Images Successfully Exported")
    for output_file in output_files:
        print(output_file)

视觉转换

from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a document.proto or sharded document.proto in path gs://bucket/path/to/folder
# gcs_bucket_name = "bucket"
# gcs_prefix = "path/to/folder"


def convert_document_to_vision_sample(
    gcs_bucket_name: str,
    gcs_prefix: str,
) -> None:
    wrapped_document = document.Document.from_gcs(
        gcs_bucket_name=gcs_bucket_name, gcs_prefix=gcs_prefix
    )

    # Converting wrapped_document to vision AnnotateFileResponse
    annotate_file_response = (
        wrapped_document.convert_document_to_annotate_file_response()
    )

    print("Document converted to AnnotateFileResponse!")
    print(
        f"Number of Pages : {len(annotate_file_response.responses[0].full_text_annotation.pages)}"
    )

hOCR 转换

from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a document.proto or sharded document.proto in path gs://bucket/path/to/folder
# document_path = "path/to/local/document.json"
# document_title = "your-document-title"


def convert_document_to_hocr_sample(document_path: str, document_title: str) -> str:
    wrapped_document = document.Document.from_document_path(document_path=document_path)

    # Converting wrapped_document to hOCR format
    hocr_string = wrapped_document.export_hocr_str(title=document_title)

    print("Document converted to hOCR!")
    return hocr_string

第三方转化

from google.cloud.documentai_toolbox import converter

# TODO(developer): Uncomment these variables before running the sample.
# This sample will convert external annotations to the Document.json format used by Document AI Workbench for training.
# To process this the external annotation must have these type of objects:
#       1) Type
#       2) Text
#       3) Bounding Box (bounding boxes must be 1 of the 3 optional types)
#
# This is the bare minimum requirement to convert the annotations but for better accuracy you will need to also have:
#       1) Document width & height
#
# Bounding Box Types:
#   Type 1:
#       bounding_box:[{"x":1,"y":2},{"x":2,"y":2},{"x":2,"y":3},{"x":1,"y":3}]
#   Type 2:
#       bounding_box:{ "Width": 1, "Height": 1, "Left": 1, "Top": 1}
#   Type 3:
#       bounding_box: [1,2,2,2,2,3,1,3]
#
#   Note: If these types are not sufficient you can propose a feature request or contribute the new type and conversion functionality.
#
# Given a folders in gcs_input_path with the following structure :
#
# gs://path/to/input/folder
#   ├──test_annotations.json
#   ├──test_config.json
#   └──test.pdf
#
# An example of the config is in sample-converter-configs/Azure/form-config.json
#
# location = "us",
# processor_id = "my_processor_id"
# gcs_input_path = "gs://path/to/input/folder"
# gcs_output_path = "gs://path/to/input/folder"


def convert_external_annotations_sample(
    location: str,
    processor_id: str,
    project_id: str,
    gcs_input_path: str,
    gcs_output_path: str,
) -> None:
    converter.convert_from_config(
        project_id=project_id,
        location=location,
        processor_id=processor_id,
        gcs_input_path=gcs_input_path,
        gcs_output_path=gcs_output_path,
    )

文档批次

from google.cloud import documentai
from google.cloud.documentai_toolbox import gcs_utilities

# TODO(developer): Uncomment these variables before running the sample.
# Given unprocessed documents in path gs://bucket/path/to/folder
# gcs_bucket_name = "bucket"
# gcs_prefix = "path/to/folder"
# batch_size = 50


def create_batches_sample(
    gcs_bucket_name: str,
    gcs_prefix: str,
    batch_size: int = 50,
) -> None:
    # Creating batches of documents for processing
    batches = gcs_utilities.create_batches(
        gcs_bucket_name=gcs_bucket_name, gcs_prefix=gcs_prefix, batch_size=batch_size
    )

    print(f"{len(batches)} batch(es) created.")
    for batch in batches:
        print(f"{len(batch.gcs_documents.documents)} files in batch.")
        print(batch.gcs_documents.documents)

        # Use as input for batch_process_documents()
        # Refer to https://cloud.google.com/document-ai/docs/send-request
        # for how to send a batch processing request
        request = documentai.BatchProcessRequest(
            name="processor_name", input_documents=batch
        )
        print(request)

合并文档分片

from google.cloud import documentai
from google.cloud.documentai_toolbox import document

# TODO(developer): Uncomment these variables before running the sample.
# Given a document.proto or sharded document.proto in path gs://bucket/path/to/folder
# gcs_bucket_name = "bucket"
# gcs_prefix = "path/to/folder"
# output_file_name = "path/to/folder/file.json"


def merge_document_shards_sample(
    gcs_bucket_name: str, gcs_prefix: str, output_file_name: str
) -> None:
    wrapped_document = document.Document.from_gcs(
        gcs_bucket_name=gcs_bucket_name, gcs_prefix=gcs_prefix
    )

    merged_document = wrapped_document.to_merged_documentai_document()

    with open(output_file_name, "w") as f:
        f.write(documentai.Document.to_json(merged_document))

    print(f"Document with {len(wrapped_document.shards)} shards successfully merged.")