Teaching resource · Department of English and ASCA
Ask one question of every text in a collection. See whether texts change over time, differ between groups, or bear out what is generally assumed.
Have American presidents become plainer spoken? Are women’s and men’s lives remembered differently in their obituaries? How did dime novels portray Native characters? The Text Annotation Lab lets you put a question like these to every text in a collection and get a structured answer for each one, ready for a spreadsheet. You write the prompt, an AI model applies it to each text in turn, and you interpret the results.
From the worked example
Are presidents getting plainer?
All 60 inaugural addresses, 1789–2025, rated for linguistic complexity from 1 (plain) to 7 (dense).
Score 7 · John Adams, 1797“When it was first perceived, in early times, that no middle course for America remained between unlimited submission to a foreign legislature and a total independence of its claims, men of reflection were less apprehensive of danger from the formidable power of fleets and armies they must determine to resist than from those contests and dissensions which would certainly arise concerning the forms of government to be instituted over the whole and over the parts of this extensive country.”
Score 3 · Franklin D. Roosevelt, 1945“We have learned to be citizens of the world, members of the human community.”
Each score rates a whole speech (one run, simple prompt). The quotations are the model’s own picks to illustrate each score, from a GPT-6 Luna run that gave both speeches the same scores. Hover over a dot to see the address.
The trend confirms the familiar story. Exceptions, like Roosevelt’s brief wartime address of 1945, can also point you to speeches worth a closer look.
Video walkthrough
Watch a mini-study from start to finish
A 3½-minute walkthrough of one mini-study, from question to results.
AI annotations are measurements to be checked, not facts. The Lab’s guide shows how to pilot a prompt and test the results against human judgment.
A step-by-step guide, ten prompting tips with a handout, and the worked example take students from a question to a first mini-study in a single class session. All 60 addresses cost about 2 cents with OpenAI’s GPT-6 Luna.
Developed for teaching by Jens Kjeldgaard-Christiansen, Department of English, Aarhus University, and the American Studies Center Aarhus, with Edgar Dubourg (Paris School of AI, PSL) as methodological advisor. Your studies are saved only in your browser; the texts you annotate are sent to the AI provider you choose.
