The Scenario
You're a content designer at an e-learning company and the curriculum review is next Thursday. You have 300 rows in your Google Sheet: column A has the reference answer for each quiz question, column B has a student response pulled from the LMS, and column C needs a 0–1 similarity score so you can flag responses that are technically answering the question but missing the key concept. Column C is empty.
The bad version:
- Try a cosine similarity formula in Sheets — realize it requires word vectors you don't have and the formula approach gives you token overlap, not semantic meaning
- Ask the engineering team to run a similarity model, get told it's not a priority this sprint
- Manually read every response pair and assign a score yourself, get to row 40, realize your scores for rows 1–10 were calibrated differently than your scores for rows 30–40 because your frame of reference drifted
The curriculum review is in a week. You need 300 scores. Your calibration is already inconsistent at row 40.
The Easy Way: One Prompt in SheetXAI
SheetXAI is an AI agent that lives inside your Google Sheet. It uses Tisane's semantic similarity engine to score each text pair and write the result directly back into your sheet.
For every row, calculate the Tisane semantic similarity score between the text in the 'Reference Answer' column and the 'Student Response' column and write the score into a new 'Similarity' column
What You Get
- A new 'Similarity' column filled with numeric scores between 0 and 1 for all 300 rows
- Scores reflect semantic overlap, not just word matching — "dogs are carnivores" and "canines primarily eat meat" score high even with no shared words
- Rows where either column is blank get a blank in 'Similarity', not a zero
What If the Data Is Not Quite Ready
Some student responses are very short — a single word or a sentence fragment
Calculate Tisane similarity scores between 'Reference Answer' and 'Student Response' for all rows — flag any row where 'Student Response' is fewer than 5 words by writing 'Too Short' in the 'Similarity' column instead of a score
The columns have extra newlines and formatting artifacts from the LMS export
Before scoring, strip leading/trailing whitespace and remove any line breaks from both the 'Reference Answer' and 'Student Response' columns, then calculate Tisane similarity and write the score into 'Similarity'
You want to highlight low-scoring rows immediately
Calculate Tisane semantic similarity between columns A and B for all 300 rows, write the score into column C, then highlight any row in red where the score in column C is below 0.4
Full pipeline: score, flag outliers, and summarize distribution in one shot
Calculate Tisane similarity scores for all rows using 'Reference Answer' and 'Student Response', write results to 'Similarity', highlight rows below 0.4 in red, and create a summary at the bottom of the sheet showing the count of rows in each band: below 0.4, 0.4–0.7, above 0.7
One prompt handles scoring, conditional formatting, and the summary breakdown together.
Try It
Get the 7-day free trial of SheetXAI and open any Google Sheet where you need to compare two text columns for meaning, then ask it to score the similarity across all rows. To extend this into a full content review pipeline, see bulk text analysis or bulk translation with Tisane. The full Tisane integration is documented in the hub.