cortex-classify-tutorial

Snowflake Cortex CLASSIFY_TEXT classifies unstructured text into discrete categories using Snowpark Python DataFrames or SQL queries. It processes single strings or batch columns and outputs JSON variant labels with confidence scores.

You classify customer reviews in Snowpark Python by applying CLASSIFY_TEXT to a DataFrame column. Define a task description and category array, then execute the operation to generate label and confidence outputs for each review.

Yes, you need a Snowflake account with Cortex AI enabled to use the CLASSIFY_TEXT function. This environment provides the necessary infrastructure to run text classification tasks directly within your data warehouse.

The best way to design categories for text classification is to define clear, discrete labels that match your routing or monitoring goals. Provide a descriptive task description and a category array so the model can accurately assign labels and confidence values.

Yes, you can create a SQL view with classified text labels by embedding the CLASSIFY_TEXT function in your view definition. This allows continuous classification of new unstructured text data as it enters your tables.

Low confidence scores in Cortex classification often occur when the task description is vague or the category array lacks distinct labels. Refine your category design and ensure the task description accurately reflects the text input.