IBM Granite

Check out this video from the Make Data Useful tutorials series:

Structured Data Extraction from Unstructured Text Python LLMs Ollama Pydantic Llama 3.2 Granite 3.2

In this tutorial, learn how to effortlessly convert unstructured text into structured data using Python and Large Language Models (LLMs). We’ll show you how to leverage a local LLM setup with Ollama, featuring Meta’s Llama 3.2 and IBM’s Granite 3.2, to extract key information from support tickets and other text data.

You'll discover how to generate clean JSON outputs from raw text and enforce data structure using Pydantic’s BaseModel and model_json_schema(). Plus, we’ll share tips on prompt engineering to improve accuracy and demonstrate how these powerful tools can streamline data cleaning and transformation.

By the end of this tutorial, you’ll know how to:

• Extract structured data from unstructured text using local LLMs
• Use Pydantic to validate AI-generated data
• Optimize data parsing with Llama 3.2 and Granite 3.2
• Apply Python techniques to enhance your data science workflow

Whether you're working with support tickets, customer messages, or other unstructured text, this guide will help you automate and optimize your data extraction process.

Links mentioned

• Ollama Software: https://ollama.com
• Python Package: https://pypi.org/project/ollama/
• Granite3.2 Model: https://ollama.com/library/granite3.2
• Llama3.2 Model: https://ollama.com/library/llama3.2

#LLM