AI System for Matching Exam Documents with Session Records

The problem

When answer sheets, candidate numbers and room lists are held in different sources, staff must match the fields manually before preparing the conflicting-document list.

The exam coordinator must compare every suggestion in the conflicting-document list with the source record; missing or conflicting evidence must not replace the final decision.

If an unreadable mark or room change is not routed for separate review, the document could be linked to the wrong student; the record must not proceed without a human decision.

What we set out to improve

• Time taken for the draft conflicting-document list and its supporting evidence to reach the exam coordinator for review
• Resolution outcomes for records routed to human review because of an unreadable mark or room change
• Number of suggestions corrected after identifying the risk of linking a document to the wrong student

How the system works

1. 01 Retrieve answer sheets, candidate numbers and room lists from authorised sources
2. 02 Verify field and document versions
3. 03 Prepare the conflicting-document list with its supporting evidence
4. 04 Check the exception: an unreadable mark or room change
5. 05 Have the exam coordinator accept, correct or reject the output
6. 06 Decision gate: document-to-student matching and grade decision
7. 07 Record the decision and source separately

Methods we used

• A source-identified data schema for answer sheets, candidate numbers and room lists
• A structure that separates deterministic rules from AI suggestions
• A screen showing the conflicting-document list alongside its supporting evidence
• A human review queue for an unreadable mark or room change
• A record identifier that prevents duplicate processing

Where people stay involved

Data and security

Only the necessary fields from answer sheets, candidate numbers and room lists should be processed. Role-based access must be applied, and unnecessary identity, health, financial and internal institutional data must not be sent to the model. Source retrieval, AI suggestions and the exam coordinator's decision must be logged separately.

Who this suits

Frequently asked questions