# Data cleaning rules — prompt for Gemma

Find inconsistencies in a small text dataset and describe transformations without losing original values.

Canonical URL: https://boswidget.ru/en/data-cleaning-rules-prompt-for-gemma/

Model: Gemma

Category: Analysis

Tags: data, cleaning, normalization

![Data cleaning rules](https://boswidget.ru/assets/images/text-preview-gemma-data-normalization-1200.3b882c3971c6.webp)

Data cleaning rules

## Prompt

Review this textual data sample: {sample}. Field meanings: {meaning}. Make an issue table: field | example | problem | proposed rule | risk of losing meaning. Distinguish safe transformations from those requiring the data owner's decision. Do not automatically merge similar records or fill gaps with averages. Give five checks to run after cleaning. Do not silently correct source values; suggest retaining them alongside normalized values.

## Template fields

- `{sample}` — Data sample. Example input: Paste a few CSV or table rows without personal information
- `{meaning}` — Meaning of the fields. Example input: What each column means and which values are permitted

## Launch

[Open Gemma in YesAI](https://yesai.su/en/gemma)