[Textract] [ExpenseAnalysis] [Asynchronous] - How to extract just summary fields?

Question

Hello, I'm trying to extract just the summary fields from a `startExpenseAnalysis` call with Textract (ignore the line item fields) , is it that possible?

I'm trying to reduce the response time less than 30 seconds since I'd like to implement it in an APIGateway endpoint.

Thanks.

Alex_T · Accepted Answer

I believe the short answer to your question is **no**: It's not possible today to specify Textract AnalyzeExpense should *only* consider the summary fields to save time. There are no parameters in either the synchronous [AnalyzeExpense](https://docs.aws.amazon.com/textract/latest/dg/API_AnalyzeExpense.html) or asynchronous [StartExpenseAnalysis](https://docs.aws.amazon.com/textract/latest/dg/API_StartExpenseAnalysis.html) to achieve this.

If you're using the async APIs, do make sure you're using the event-driven SNS call-back rather than polling the job status (which introduces poll wait delay).

If your workload is small/low-concurrency (no or minimal parallel requests), you coooooould explore splitting input documents yourself and using the synchronous AnalyzeExpense API: But this approach wouldn't scale well due to the [Textract Quota Limits](https://docs.aws.amazon.com/textract/latest/dg/limits-quotas-explained.html) - and I wouldn't guarantee without testing whether it'd be faster than async anyway.

Like the other answer mentioned, you could explore trying to optimize your documents themselves for faster processing?

AWS TAM · Answer

To extract just the summary fields from a startExpenseAnalysis call with Textract, you can follow these steps:

1. Use the `StartExpenseAnalysis` API to initiate the asynchronous analysis of the invoices or receipts stored in an Amazon S3 bucket. This API will return a `JobId` that you can use to retrieve the results later.

2. Use the `GetExpenseAnalysis` API to retrieve the results of the expense analysis operation. The response will contain a `SummaryFields` section that includes the extracted summary information, such as the total amount, currency, and other high-level details.

3. To reduce the response time to less than 30 seconds, you can consider the following approaches:

- Optimize the input documents: Ensure that the invoices or receipts are in a format (JPEG, PNG, or PDF) that Textract can process efficiently. Also, make sure the documents are of good quality and not too large.

- Use asynchronous processing: As mentioned, the `StartExpenseAnalysis` API initiates an asynchronous operation. This allows you to retrieve the results later using the `GetExpenseAnalysis` API, which should be faster than a synchronous operation.

- Implement caching: If you are processing the same documents repeatedly, you can cache the summary results to avoid re-processing the documents every time.

- Optimize your API Gateway endpoint: Ensure that your API Gateway endpoint is configured correctly, with appropriate caching, throttling, and other performance-related settings.

4. Note that the `SummaryFields` section of the `GetExpenseAnalysis` response will only contain the high-level summary information, and not the line item details. If you need to extract the line item information as well, you can access the `LineItemGroups` section of the response.

[StartExpenseAnalysis - Amazon Textract](https://docs.aws.amazon.com/textract/latest/dg/API_StartExpenseAnalysis.html)

[AnalyzeExpense - Amazon Textract](https://docs.aws.amazon.com/textract/latest/dg/API_AnalyzeExpense.html)

[GetExpenseAnalysis - Amazon Textract](https://docs.aws.amazon.com/textract/latest/dg/API_GetExpenseAnalysis.html)

[Textract] [ExpenseAnalysis] [Asynchronous] - How to extract just summary fields?

Relevant content