通过利用无服务器技术，我们能够构建灵活、高效且经济实惠的转录与翻译解决方案，帮助企业和个人在多语言环境中实现无缝交流。本文将探讨如何利用简单的事件驱动无服务器架构来使用Amazon Translate和Amazon Transcribe，并创建一个小型编排以实现音频文件的转录和翻译。
一、架构概述
当音频文件以 “/audio” 前缀上传至 Amazon Simple Storage Service (AWS S3) 存储桶时，将触发一个 AWS Lambda 函数，该函数将利用 Amazon Transcribe 从音频文件中提取文本。转录后的文本将被保存到同一存储桶的 “/transcriptions” 目录下。
当 “/transcriptions” 目录中创建新文件时，另一个 AWS Lambda 函数将被触发以执行翻译任务。为简化说明，我们假设录音为英语，并希望将其翻译为意大利语。这个 Lambda 函数会将翻译结果写入 “/translations” 目录。
此编排非常适合用于创建小型原型和概念验证。在本例中，我们将 Amazon S3 作为消息总线。若将该架构推广到生产环境，建议使用更复杂的编排工具，如 AWS Step Functions。
注意事项：
1、使用 Amazon S3、S3 前缀和 AWS Lambda 实现编排时，请务必小心，以免导致 Lambda 函数出现递归调用，这样会产生不必要的费用。例如上面提到的两个函数之一误写入了 “/audio” 目录，就会陷入无限循环！为了防止这种情况，建议限制函数的调用频率，并可能为所有的编排工具添加紧急停止开关。一种常见策略是为每个编排工具提供一个 Amazon S3 路径的拒绝列表；
2、如果在触发函数的事件中检测到上述路径之一，可以对该函数进行调用频率限制或忽略该事件，甚至将事件转发到死信队列并向通知系统发送警报。有些用户可能更喜欢为每个阶段保留独立的存储桶，但请注意 AWS 账户限额。如需更多存储桶，可申请增加服务配额。
二、创建AWS CDK堆栈
1、安装AWS CDK CLI
如果还未安装 AWS CDK，请使用 npm 进行安装：
npm install -g aws-cdk
2、新建AWS CDK项目
为 AWS CDK 项目新建一个目录，并初始化项目：
mkdir transcribe-translate cd transcribe-translate cdk init app --language typescript
3、添加所需的AWS CDK依赖项
运行以下命令安装必要的依赖：
npm install @aws-cdk/aws-s3 @aws-cdk/aws-lambda @aws-cdk/aws-iam @aws-cdk/aws-s3-notifications @aws-cdk/aws-lambda-nodejs
二、创建基础架构
编辑 “lib/transcribe-translate-stack.ts” 文件以定义资源：
<div class="lb-code lb-ctxt-p lb-no-highlighting" data-lb-comp="code" data-lb-comp-registered="true"><pre class=" line-numbers language-plainText"><code class="language- language-plainText">import * as cdk from 'aws-cdk-lib';
import { Construct } from 'constructs';
import * as s3 from 'aws-cdk-lib/aws-s3';
import * as lambda from 'aws-cdk-lib/aws-lambda';
import * as s3n from 'aws-cdk-lib/aws-s3-notifications';
import { NodejsFunction } from 'aws-cdk-lib/aws-lambda-nodejs';
import * as iam from 'aws-cdk-lib/aws-iam';
import * as path from 'path';
export class TranscribeTranslateStack extends cdk.Stack {
constructor(scope: Construct, id: string, props?: cdk.StackProps) {
super(scope, id, props);
// Create the S3 bucket
const bucket = new s3.Bucket(this, 'UploadBucket', {
removalPolicy: cdk.RemovalPolicy.DESTROY, // this is OK only for dev
autoDeleteObjects: true, // this is OK only for dev
});
// Create the Transcribe Lambda function
const transcribeFunction = new NodejsFunction(this, 'TranscribeFunction', {
runtime: lambda.Runtime.NODEJS_18_X,
entry: path.join(__dirname, '../lambda/transcribe.mjs'),
handler: 'handler',
environment: {
BUCKET_NAME: bucket.bucketName,
},
timeout: cdk.Duration.seconds(900),
});
// Grant read/write permissions on S3 objects for the transcribe function
bucket.grantReadWrite(transcribeFunction, 'transcriptions/*');
// Additional permissions for Transcribe
transcribeFunction.addToRolePolicy(new iam.PolicyStatement({
actions: ['transcribe:StartTranscriptionJob', 'transcribe:GetTranscriptionJob'],
resources: ['*'],
}));
// Add S3 event notification to trigger Transcribe function
bucket.addEventNotification(s3.EventType.OBJECT_CREATED, new s3n.LambdaDestination(transcribeFunction), {
prefix: 'audio/',
});
// Create the Translate Lambda function
const translateFunction = new NodejsFunction(this, 'TranslateFunction', {
runtime: lambda.Runtime.NODEJS_18_X,
entry: path.join(__dirname, '../lambda/translate.mjs'),
handler: 'handler',
environment: {
BUCKET_NAME: bucket.bucketName,
},
timeout: cdk.Duration.seconds(900),
});
// Grant read/write permissions on specific S3 objects for the translate function
bucket.grantRead(translateFunction, 'transcriptions/*');
bucket.grantReadWrite(translateFunction,'translations/*');
// Additional permissions for Translate
translateFunction.addToRolePolicy(new iam.PolicyStatement({
actions: ['translate:TranslateText'],
resources: ['*'],
}));
// Add S3 event notification to trigger Translate function
bucket.addEventNotification(s3.EventType.OBJECT_CREATED, new s3n.LambdaDestination(translateFunction), {
prefix: 'transcriptions/',
});
}
}
<button class="lb-code-copy-btn"><div class="lb-code-copy-confirmation"><i class="icon-check">复制<hr class="lb-none-v-margin lb-divider-light"><div id="zlrw"><h2 id=".E8.A7.A6.E5.8F.91.E8.BD.AC.E5.BD.95.E4.BB.BB.E5.8A.A1" class="lb-txt-none lb-none-v-margin lb-h2 lb-title">触发转录任务<div class="lb-txt-16 lb-none-v-margin lb-rtxt">
在以下示例中，我们将触发转录任务，但不会监控其是否成功完成。这种乐观方法只有在创建原型时才会有效！如果投产，建议是轮询 <a href=”https://docs.aws.amazon.com/AWSJavaScriptSDK/v3/latest/client/transcribe/command/GetTranscriptionJobCommand/”>Amazon Transcribe API ，以便您知道任务何时成功完成。
在本地代码仓库创建一个 lambda/transcribe.mjs 文件
import * as cdk from 'aws-cdk-lib';
import { Construct } from 'constructs';
import * as s3 from 'aws-cdk-lib/aws-s3';
import * as lambda from 'aws-cdk-lib/aws-lambda';
import * as s3n from 'aws-cdk-lib/aws-s3-notifications';
import { NodejsFunction } from 'aws-cdk-lib/aws-lambda-nodejs';
import * as iam from 'aws-cdk-lib/aws-iam';
import * as path from 'path';
export class TranscribeTranslateStack extends cdk.Stack {
constructor(scope: Construct, id: string, props?: cdk.StackProps) {
super(scope, id, props);
// Create the S3 bucket
const bucket = new s3.Bucket(this, 'UploadBucket', {
removalPolicy: cdk.RemovalPolicy.DESTROY, // this is OK only for dev
autoDeleteObjects: true, // this is OK only for dev
});
// Create the Transcribe Lambda function
const transcribeFunction = new NodejsFunction(this, 'TranscribeFunction', {
runtime: lambda.Runtime.NODEJS_18_X,
entry: path.join(__dirname, '../lambda/transcribe.mjs'),
handler: 'handler',
environment: {
BUCKET_NAME: bucket.bucketName,
},
timeout: cdk.Duration.seconds(900),
});
// Grant read/write permissions on S3 objects for the transcribe function
bucket.grantReadWrite(transcribeFunction, 'transcriptions/*');
// Additional permissions for Transcribe
transcribeFunction.addToRolePolicy(new iam.PolicyStatement({
actions: ['transcribe:StartTranscriptionJob', 'transcribe:GetTranscriptionJob'],
resources: ['*'],
}));
// Add S3 event notification to trigger Transcribe function
bucket.addEventNotification(s3.EventType.OBJECT_CREATED, new s3n.LambdaDestination(transcribeFunction), {
prefix: 'audio/',
});
// Create the Translate Lambda function
const translateFunction = new NodejsFunction(this, 'TranslateFunction', {
runtime: lambda.Runtime.NODEJS_18_X,
entry: path.join(__dirname, '../lambda/translate.mjs'),
handler: 'handler',
environment: {
BUCKET_NAME: bucket.bucketName,
},
timeout: cdk.Duration.seconds(900),
});
// Grant read/write permissions on specific S3 objects for the translate function
bucket.grantRead(translateFunction, 'transcriptions/*');
bucket.grantReadWrite(translateFunction,'translations/*');
// Additional permissions for Translate
translateFunction.addToRolePolicy(new iam.PolicyStatement({
actions: ['translate:TranslateText'],
resources: ['*'],
}));
// Add S3 event notification to trigger Translate function
bucket.addEventNotification(s3.EventType.OBJECT_CREATED, new s3n.LambdaDestination(translateFunction), {
prefix: 'transcriptions/',
});
}
}
三、触发转录任务
在本地代码仓库创建一个 “lambda/transcribe.mjs” 文件，其中包含转录任务的实现代码。该任务完成后，将在 “/transcriptions” 目录下新建一个对象，触发后续翻译任务的编排。
import { TranscribeClient, StartTranscriptionJobCommand } from '@aws-sdk/client-transcribe';
const transcribeClient = new TranscribeClient();
export const handler = async (event) => {
const bucket = process.env.BUCKET_NAME;
const key = event.Records[0].s3.object.key;
const transcribeParams = {
TranscriptionJobName: `TranscribeJob-${Date.now()}`,
LanguageCode: 'en-US',
Media: {
MediaFileUri: `s3://${bucket}/${key}`
},
OutputBucketName: bucket,
OutputKey: `transcriptions/${key.split("/").pop()}.json`
};
try {
await transcribeClient.send(new StartTranscriptionJobCommand(transcribeParams));
console.log('Transcription job started');
} catch (err) {
console.error('Error starting transcription job', err);
return;
}
};
四、触发翻译任务
创建一个 “lambda/translate.mjs” 文件，编写翻译任务的实现代码。
import { S3Client, GetObjectCommand, PutObjectCommand } from '@aws-sdk/client-s3';
import { TranslateClient, TranslateTextCommand } from '@aws-sdk/client-translate';
const s3Client = new S3Client();
const translateClient = new TranslateClient();
export const handler = async (event) => {
const bucket = process.env.BUCKET_NAME;
const key = event.Records[0].s3.object.key;
// Get the transcription text from S3
let transcriptionData;
try {
transcriptionData = await s3Client.send(new GetObjectCommand({
Bucket: bucket,
Key: key
}));
} catch (err) {
console.error('Error getting transcription from S3', err);
return;
}
const { results } = JSON.parse(await streamToString(transcriptionData.Body));
const transcriptionText = results.transcripts.reduce(
(acc, curr) => acc + curr.transcript, ''
);
console.log('Transcription text:', transcriptionText);
// Translate the transcription
const translateParams = {
SourceLanguageCode: 'en',
TargetLanguageCode: 'it',
Text: transcriptionText
};
console.log(translateParams);
let translation;
try{
translation = await translateClient.send(new TranslateTextCommand(translateParams));
console.log('Translation:', translation.TranslatedText);
}catch(err){
console.error('Error translating text', err);
return;
}
// Save the translation result to S3
const translationResult = {
originalText: transcriptionText,
translatedText: translation.TranslatedText
};
try{
await s3Client.send(new PutObjectCommand({
Bucket: bucket,
Key: `translations/${key.split("/").pop()}.it.txt`,
Body: translationResult.translatedText,
ContentType: 'text/plain'
}));
}catch(err){
console.error('Error saving translation to S3', err);
return;
}
console.log('Translation saved');
};
// Helper function to convert a stream to a string
const streamToString = (stream) => {
return new Promise((resolve, reject) => {
const chunks = [];
stream.on('data', (chunk) => chunks.push(chunk));
stream.on('error', reject);
stream.on('end', () => resolve(Buffer.concat(chunks).toString('utf8')));
});
};
五、部署
运行以下命令进行部署：
cdk bootstrap # if you haven't bootstrapped in this region before cdk deploy
完成部署后，应该能看到一个输出列表，其中包含 Amazon S3 存储桶名称等信息。
六、测试
上传 MP3 文件到 “audio/” 目录以进行测试。可以使用以下 AWS CLI 命令上传文件：
aws s3 cp ./path/to/your/audio/file s3://your-bucket-here/audio/file.mp3
或者访问 AWS 管理控制台，将 MP3 文件直接上传到 “audio/”。稍后，将在 “transcriptions/” 和 “translations/” 目录中找到每个类别的编排结果。
七、清理
任务完成后，可以通过以下命令拆解基础架构：
cdk destroy
相关推荐：
-
广告合作
-
QQ群号：707632017
