Batch Analysis — Uploading, Analyzing and Batch Marking

This guide shows how to use the UnaGPT Batch Analyze Wizard to upload a question paper, have the AI analyze the questions and the marking scheme, mark a whole class of student scripts in one batch, and review the scores and reports. Every step comes with a screenshot; the red numbered markers in the screenshots correspond to the numbered markers in the text.

• Platform: app.una.study
• Audience: Teachers
• The Batch Analyze Wizard is currently in public beta (BETA).

Part 1 — Create a project and upload the paper

Open https://app.una.study/ and choose "Log in using a third-party account" (marker 1) to sign in with the school Microsoft or Google account.

After logging in, click "Batch Analysis" in the left menu (marker 1) to open the wizard.

On the "Projects" page, click "New project" (marker 1). Each project holds one question paper, its marking scheme and the student scripts marked against it. The "Status" column (marker 2) shows the stage each project has reached: Draft, Analyzing, Analyzed, Marking, Complete.

Type the project title in the dialog (marker 1, for example: F6 ICT Mock Paper), then click "Create" (marker 2).

Choose how the material is organized at the top right (marker 1): "Separate documents" or "One combined document". One file holding both documents goes into "A stitched scan" and is split automatically (marker 2); otherwise the paper goes into "Question paper" (marker 3) and the scheme into "Marking scheme" below it (marker 4). Files can be dragged onto an area, or the area can be clicked to browse.

Once a file is uploaded the system generates page previews. Use "All" or "None" for a quick selection (marker 1) and confirm with "Choose N pages" (marker 2); the ticked thumbnails (marker 3) are the pages that will be analyzed.

The marking scheme is confirmed the same way (marker 1), or it can be left for the AI to generate in the next step.

When both documents show "Ready" (marker 1), check the page thumbnails once more, then click "Analyze · N pages" at the bottom of the page (marker 2).

The system detects the questions, parses the marking scheme and builds the question bounds, showing its progress as it goes (marker 1).

When the analysis finishes, the number of questions detected and the total marks are shown (marker 1). Click "Review result" to open the review page (marker 2); if the detection looks wrong, "Re-analyze" runs it again.

Part 2 — Review the questions and the marking scheme

The top of the review page shows the number of questions detected and the total marks (marker 1). Questions with no marking scheme can be filled in by the AI with "Generate marking scheme" (marker 2). "Answer language" is detected from the paper — marking and reports are written in this language, so correct it here if it is wrong (marker 4). When everything is correct, click "Start marking" (marker 3).

Questions that have parts show their (a), (b), (c) structure. "Edit" opens the question text for changes (marker 1). Each part carries its own marks and question type (marker 2), and the handle on the right (marker 3) opens that part's content and marks.

Part 3 — Mark the student scripts

The marking page offers two ways to upload: one file per student, dropped on "Add more student scripts" (marker 1); or one scan of the whole class, dropped on the right (marker 2), which the system splits per student.

A class scan is sorted into one script per student. Check the name on each script and correct it if needed; "Sort again" repeats the split. Click "Upload N scripts · N pages" to confirm.

The system reads each script in turn — OCR for handwriting — and marks it against the marking scheme. The status chip and the counter beside it show overall progress (marker 1), the row below names the script being processed (marker 2), and the queue can be paused or stopped there (marker 3).

Part 4 — Scores and reports

When marking finishes the status reads "Complete", with the number of scripts marked beside it (marker 1).

The "Class overview" tab shows the average, highest and lowest scores, the pass rate and the scans to check (marker 1); three charts — score distribution, score rate by question part and grade distribution (marker 2); and the class list below (marker 3), with each question's marks, the total, the grade and the review status.

The "Review" tab compares the whole class question by question. Questions are switched on the left (marker 1). The student's handwriting (marker 2) sits beside the text the system read from it (marker 3); if the recognition is wrong, use "Edit" or run "OCR" again. "View original" opens the student's whole script (marker 4).

The AI awards marks criterion by criterion and gives its reasons (marker 1). Each criterion can be adjusted in the right-hand column (marker 2), with a note explaining the change (marker 3). Adjusted scores sync to the class overview and the reports, and the student is flagged as teacher-adjusted.

The "Reports" tab builds a detailed report for each student. Students are switched with the dropdown (marker 1), the teacher and student versions with the toggle (marker 2), and "Print PDF" exports the report (marker 3). The header shows the total, the percentage and the grade (marker 4).

The student version (marker 1) gives every question two views, a score report and a content analysis (marker 2).

"Grade settings" on the class overview sets the pass mark as a percentage (marker 1) and the band system — S–E, HKDSE or Pass/Fail — where each band's name and threshold can be edited or a band added (marker 2). "Save settings" applies the change (marker 3) and the overview and the reports update immediately.

The "Download" menu (marker 1) offers three exports (marker 2): CSV, the class score table; All-in-one PDF, every report in one file; and Individual PDFs (ZIP), one file per student.

The "Re-mark" menu (marker 1) offers "Re-mark all", which re-scores the class with the current marking scheme, and "Re-extract & mark all", which reads the scripts again first (marker 2). If the marking scheme has changed, update it in the review page before re-marking.