Author(s): Júlio Almeida

Originally published on Towards AI.

Extracting structured data from PDFs and images can be challenging, but combining Optical Character Recognition (OCR) with Language Models (LLMs) offers a powerful solution. Within the Azure ecosystem, Azure Document Intelligence is the way to go when analyzing documents.

In this article, I will demonstrate how to leverage the Phi-3 mini model from the Azure AI studio to enhance the data extraction process. The Phi-3 mini model, a small language model (SML) with 3.8 billion parameters, provides efficient and accurate results, making it an ideal choice for this task. While this example focuses on Azure, the principles can be applied using similar tools from other providers.

The solution can be easily replicated for other cloud providers with comparable quality, pricing, and features.

Azure Document Intelligence is an OCR product present in the Microsoft ecosystem. In terms of pricing, it is divided into three tiers as you can see in the image below:

Prices for each DocumentType

The “read” only extracts paragraphs and handwritten text. Essentially a pure traditional OCR. This would be enough for most of the document extraction, but for complex documents will not suffice.

The “prebuilt” layout is the best choice for the job, the rest will be done by the LLM. This option… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI