S3 に保存した画像から Amazon Textract でテキストデータを取得する

npm init
npm install aws-sdk

import AWS from 'aws-sdk';

const textract = new AWS.Textract();
const s3 = new AWS.S3();

export const detectText = async (event) => {
  // S3 のイベントからオブジェクトの情報を取得
  const bucket = event.Records[0].s3.bucket.name;
  const objectKey = event.Records[0].s3.object.key;

  const params = {
    DocumentLocation: {
      S3Object: {
        Bucket: bucket,
        Name: objectKey
      }
    }
  };

  try {
    // Textractを呼び出し、テキストを抽出
    const response = await textract.startDocumentTextDetection(params).promise();
    
    // Textractのジョブが完了するのを待つ
    const jobId = response.JobId;
    await waitForJobCompletion(jobId);

    // テキストを取得
    const text = await getTextractText(jobId);
    console.log(text);
    
    // テキストを使用して必要な処理を行う
    const putparams = {
      Bucket: '20231028-text-bucket',
      Key: 'sample.txt',
      Body: text,
      ContentType: 'text/plain',
    };

    await s3.putObject(putparams).promise();

    return {
      statusCode: 200,
      body: 'Text file uploaded successfully.'
    };
  } catch (error) {
    return {
      statusCode: 500,
      body: JSON.stringify(error.message)
    };
  }
};

// Textractのジョブが完了するのを待つ
async function waitForJobCompletion(jobId) {
  const delay = (ms) => new Promise(resolve => setTimeout(resolve, ms));

  while (true) {
    const response = await textract.getDocumentTextDetection({ JobId: jobId }).promise();

    if (response.JobStatus === 'SUCCEEDED') {
      break;
    }

    if (response.JobStatus === 'IN_PROGRESS') {
      await delay(5000);
    }
  }
}

// Textractからテキストを取得
async function getTextractText(jobId) {
  const response = await textract.getDocumentTextDetection({ JobId: jobId }).promise();
  const text = response.Blocks
    .filter(block => block.BlockType === 'LINE')
    .map(block => block.Text)
    .join('\n');
  
  return text;
}

{
   "JobId": "string"
}

// Textractのジョブが完了するのを待つ
async function waitForJobCompletion(jobId) {
  const delay = (ms) => new Promise(resolve => setTimeout(resolve, ms));

  while (true) {
    const response = await textract.getDocumentTextDetection({ JobId: jobId }).promise();

    if (response.JobStatus === 'SUCCEEDED') {
      break;
    }

    if (response.JobStatus === 'IN_PROGRESS') {
      await delay(5000);
    }
  }
}

{
   "Blocks": [ 
      { 
         "BlockType": "string",
         "ColumnIndex": number,
         "ColumnSpan": number,
         "Confidence": number,
         "EntityTypes": [ "string" ],
         "Geometry": { 
            "BoundingBox": { 
               "Height": number,
               "Left": number,
               "Top": number,
               "Width": number
            },
            "Polygon": [ 
               { 
                  "X": number,
                  "Y": number
               }
            ]
         },
         "Id": "string",
         "Page": number,
         "Query": { 
            "Alias": "string",
            "Pages": [ "string" ],
            "Text": "string"
         },
         "Relationships": [ 
            { 
               "Ids": [ "string" ],
               "Type": "string"
            }
         ],
         "RowIndex": number,
         "RowSpan": number,
         "SelectionStatus": "string",
         "Text": "string",
         "TextType": "string"
      }
   ],
   "DetectDocumentTextModelVersion": "string",
   "DocumentMetadata": { 
      "Pages": number
   },
   "JobStatus": "string",
   "NextToken": "string",
   "StatusMessage": "string",
   "Warnings": [ 
      { 
         "ErrorCode": "string",
         "Pages": [ number ]
      }
   ]
}

// Textractからテキストを取得
async function getTextractText(jobId) {
  const response = await textract.getDocumentTextDetection({ JobId: jobId }).promise();
  console.log(JSON.stringify(response.Blocks));
  const text = response.Blocks
    .filter(block => block.BlockType === 'LINE')
    .map(block => block.Text)
    .join('\n');
  
  return text;
}