BERT Summarizer

Unstable API 0.7.0 @project-lakechain/bert-extractive-summarizer

---

The BERT extractive summarizer is a middleware allowing to summarize text documents using the BERT Extractive Summarizer and the HuggingFace Pytorch transformers libraries. It makes it possible to run extractive summarization on text documents within a document processing pipeline.

---

📝 Summarizing Text

To use this middleware, you import it in your CDK stack and connect it to a data source that provides text documents, such as the S3 Trigger if your text documents are stored in S3.

ℹ️ The below example shows how to create a pipeline that summarizes text documents uploaded to an S3 bucket.

import { S3EventTrigger } from '@project-lakechain/s3-event-trigger';
import { BertExtractiveSummarizer } from '@project-lakechain/bert-extractive-summarizer';
import { CacheStorage } from '@project-lakechain/core';

class Stack extends cdk.Stack {
  constructor(scope: cdk.Construct, id: string) {
    // Sample VPC.
    const vpc = new ec2.Vpc(this, 'VPC', {});

    // The cache storage.
    const cache = new CacheStorage(this, 'Cache');

    // Create the S3 event trigger.
    const trigger = new S3EventTrigger.Builder()
      .withScope(this)
      .withIdentifier('Trigger')
      .withCacheStorage(cache)
      .withBucket(bucket)
      .build();

    // Summarize uploaded text documents.
    trigger.pipe(new BertExtractiveSummarizer.Builder()
      .withScope(this)
      .withIdentifier('BertSummarizer')
      .withCacheStorage(cache)
      .withSource(trigger)
      .withVpc(vpc)
      .build());
  }
}

---

Summarization Ratio

You can specify a ratio indicating how much of the original text you want to keep in the summary.

💁 The default value is 0.2 (20% of the original text), and you can opt for a value between 0.1 and 1.0.

const summarizer = new BertExtractiveSummarizer.Builder()
  .withScope(this)
  .withIdentifier('BertSummarizer')
  .withCacheStorage(cache)
  .withSource(source)
  .withVpc(vpc)
  .withRatio(0.3)
  .build();

---

Auto-Scaling

The cluster of containers deployed by this middleware will auto-scale based on the number of text documents that need to be processed. The cluster scales up to a maximum of 5 instances by default, and scales down to zero when there are no documents to process.

ℹ️ You can configure the maximum amount of instances that the cluster can auto-scale to by using the withMaxInstances method. Note that this is only valid when using this middleware using a GPU compute type.

const summarizer = new BertExtractiveSummarizer.Builder()
  .withScope(this)
  .withIdentifier('BertSummarizer')
  .withCacheStorage(cache)
  .withSource(source)
  .withVpc(vpc)
  .withMaxInstances(10)
  .build();

---

🏗️ Architecture

This middleware supports both CPU and GPU compute types. We implemented 2 different architectures, one that’s GPU based and using ECS, the other which is CPU based and serverless, based on AWS Lambda. You can use the .withComputeType API to select the compute type you want to use.

💁 By default, this implementation will run on the CPU compute type.

GPU Architecture

The GPU architecture leverages AWS ECS to run the summarization process on g4dn.xlarge instances. The GPU instance is part of a ECS cluster, and the cluster is part of a VPC, running within its private subnets.

CPU Architecture

The CPU architecture leverages AWS Lambda to run the summarization process on serverless compute. The Lambda function runs as part of a VPC and is integrated with AWS EFS to cache the BERT model(s).

---

🏷️ Properties

Supported Inputs

Mime Type	Description
text/plain	UTF-8 text documents.

Supported Outputs

Mime Type	Description
text/plain	UTF-8 text documents.

Supported Compute Types

Type	Description
CPU	This middleware supports CPU compute.
GPU	This middleware supports GPU compute.

---

📖 Examples

• Extractive Summarization Pipeline - Builds a pipeline for text summarization using BERT extractive summarizer.