Implement information extraction solutions 5.2 Extract content from documents

A Document Intelligence v4.0 custom classification model routes uploaded PDFs.

Study guide skill: Extract information by using multimodal pipelines that combine OCR, layout analysis, and field extraction

Question B7-026

A Document Intelligence v4.0 custom classification model routes uploaded PDFs. Each PDF can contain an application form followed by several pay slips, with page ranges that vary. The analyze request doesn't set splitMode, and the service returns one class for the whole file. What should you set?

Select one answer.

  • A splitMode set to none, which splits files that contain several documents
  • B Leave splitMode unset, because the default already splits multi-document files
  • C splitMode set to auto, so the service identifies each document and its page range ✓ Correct answer
  • D splitMode set to perPage, so each page is classified as its own document

Answer: C

In the v4.0 API, the custom classifier doesn't split files by default; splitMode defaults to none, which returns one class for the whole file. perPage classifies each page as its own document, which breaks up multipage documents. auto identifies the documents in the file and their page ranges.

Deep dive

How it works

A Document Intelligence custom classification model can find several documents inside one input file. Since the v4.0 2024-11-30 (GA) API, the analyze operation has a splitMode property that controls this, and it doesn't split by default:

  • none (the default): the entire input file is treated as a single document and gets one class. That's what's happening now.
  • perPage: each page of the file is classified on its own.
  • auto: the service identifies the documents in the file and their page ranges.

An application form followed by a variable number of multi-page pay slips needs the service to find document boundaries itself, so set splitMode to auto. Microsoft also notes that you must explicitly set auto to keep the splitting behavior of earlier releases, which is a common migration surprise.

The other options, and when you'd use them

  • splitMode none: It treats the whole file as one document, which is the current behavior, not a fix.
  • Leave it unset: The default is none, so nothing changes.
  • splitMode perPage: Every page becomes its own classification, so a multi-page pay slip or form would be broken up instead of returned as one document with a page range.

Exam tip

v4.0 classifier: default none (one class per file). Multiple documents of varying length = auto. One class per page = perPage.

References

Last reviewed:

Ready to test yourself? The quiz asks this and every other AI-103 question, with instant feedback. Practice this in a quiz →

All Information extraction questions · AI-103 study guide

Thanks for your feedback!