Practice for language teachers

Correcting class tests and written exams: a clear routine

By Tuba ·

The first stack of papers after the start of term is a good moment to settle on a repeatable routine. Class test, written exam or quiz: this guide shows one possible workflow for written student texts, from the criteria to the feedback. You can use it with or without AI.

1. Set the criteria before the first paper

Keep the task, the learning goal and the assessment criteria at hand. For a narrative text, content, structure, vocabulary and accuracy can be separate criteria, for example. Write down what you expect under each criterion. If you use points, check that the individual values add up to the total. Use the same criteria for all papers on this task.

Practical: keep a short note on borderline cases. Later you can check whether you assessed comparable performances comparably.

2. Read the whole text first

Get an overview before you weigh individual errors: is the task answered? Is the line of thought clear? Which strength stands out? A paper with language errors can succeed in content; a formally clean text can miss the task. Separate these observations in your notes before you phrase an overall assessment.

Practical: first note one strength and one step for developing the content. The detailed correction follows afterwards.

3. Prepare photos and the task

If you work with photos, photograph each page completely, as straight as possible and without shadows. Check order and legibility. Remove names and other personal details before uploading and clarify whether using the tool is permitted at your school. Several pages of one paper belong together; papers by different people should be processed separately.

In Tuba: upload the pages of a single paper and add the year group, the task and your criteria. The app shows the cost before you start.

4. Check the transcription first, then the correction

Misread handwriting can produce an error the writer never made. So compare conspicuous passages with the photo. Then check every relevant correction: is the original form really wrong or just phrased differently? Does the explanation fit the sentence? An AI can overlook context and change correct phrasing unnecessarily.

Practical: separate real errors, pure style suggestions and uncertainly read passages. Only adopt a mark once you can trace it in the original.

5. Phrase feedback with a next step

A long list of errors does not yet say how the next paper gets better. Pick up a concrete strength, name a recurring pattern and suggest a matching exercise. Adapt the language of the feedback to the group. Automatically generated feedback is a draft that you check as a subject expert and align with the actual paper.

Practical: instead of "improve your grammar", try "Mark all verbs in your narrative and check whether they fit the time frame of the story."

6. Compare borderline cases and finalise the assessment

Reread papers with similar performance against the same criteria. Check that deductions are comprehensible and your feedback is consistent. Only then set the assessment. Tuba can prepare corrections and score suggestions; the subject-related and pedagogical decision stays with you. A PDF or printout then helps when handing the papers back.

Practical: for the next lesson, note a frequent error pattern from the stack without creating a ranking of named students.

Example: from the marked error to helpful feedback

This invented short example illustrates the routine; it is not a complete assessment.

Original text

Last summer we go to the lake. My brother swim every day.

Suggested correction

Last summer we went to the lake. My brother swam every day.

“Last summer” sets the time frame in the past. In the affirmative past simple, “go” becomes “went” and “swim” becomes “swam”.

Possible feedback

You make clear when and where the experience took place. When narrating, pay attention to the past forms. Underline the verbs and write the matching past simple form next to each. Then revise the two sentences.

Checklist before handing back

  • Task and criteria are taken into account.
  • Uncertain transcription passages were checked against the original.
  • Errors and style suggestions are kept apart.
  • Points and reasons match.
  • The feedback contains one strength and one concrete next step.
  • The final assessment was checked and set by you.

The checklist as a text template also contains fields for criteria, observations and your feedback. You can download it without an account and print it from a text editor.

Three borderline cases that need a second look

The AI recognises a different word. The original says “went”, the transcription contains “want”. Compare the letters in the photo before you derive a tense error from it. If the original allows no clear reading, flag the passage for your own check.

A correct sentence gets a more elegant phrasing. “It was a nice day” can be simple but correct. A suggestion like “It was a wonderful day” is not a required correction unless the task demands it. Keep accuracy apart from a goal such as varied expression.

A pattern repeats. Two wrong verb forms after “did” are two passages but one shared practice focus. Record how you treat recurring errors in your criteria and apply it the same way across the stack.

Try the routine with Tuba

Start with a short text you have already read yourself. That way you can judge recognition, suggested corrections and explanations before you process more papers. To get started there are 12 free red pens: four base corrections or two corrections with all modules, each with up to two pages per paper.

Test the routine on your next paper.

Upload a photo, review the suggestions and prepare feedback. Start free, no credit card.