Custom Document Field Extraction with Generative AI

The adoption of generative AI technologies is accelerating, offering transformative potential across a range of industries. Azure AI Make it possible Empowering organizations to create AI solutions that interact and respond Customize to suit your needs Playoccupy an important part helping Business harness Generative AI is effective. The new Custom Field Extraction Preview allows you to: influence Support document automation workflows by efficiently extracting fields from documents using generative AI, ensuring standardized output and repeatable processes.

Field extraction using large-scale language models

To extract fields from a document, use: Large-scale language models (LLM (Master of Laws)) or generative AI, you Typically, you will need to create complex orchestration workflows to manage multiple services, as shown below. Tasks like text extraction, document chunking, vectorization, search index creation, and prompt engineering.

• Size and complexity of prompts: Managing prompts as they change can be difficult, leading to increased number of prompts and associated costs.
• Inconsistent results: Running the same document multiple times may result in different results, which can lead to reliability issues.
• Grounding: Ensures that values ​​are extracted accurately and traceable to address hallucination issues.
• Lack of confidence scores: Without confidence scores, it is difficult to automate downstream processes.

Imagine being able to leverage the benefits of generative AI without having to develop complex workflows yourself. With the new custom field extraction feature, you simply define your schema, let the model extract the fields you need, and then fix the prediction errors. Once the model is trained, you can integrate it into your document processing workflow with a single API call. This approach provides safeguards to ensure that the extracted values ​​match your business requirements by providing evidence-based results and confidence scores.

Azure AI Document Intelligence

Azure AI Document Intelligence is an AI service that provides a streamlined set of APIs and a studio experience to efficiently extract content, structure (e.g., tables, paragraphs, sections, figures), and fields, whether predefined for a specific document type or custom for all documents or forms. With Document Intelligence API, you can easily segment, classify, and extract fields or content at scale from any document or form, tailoring it to your business needs. recent Extract document fields model Leverage Generative AI extracts user-specified fields from documents using a variety of visual templates.~is Custom Extraction This model combines the power of document understanding, large-scale language models (LLMs), and the rigor and schema of custom extraction features to produce highly accurate models in minutes.

Why should I choose Azure Document Field Extraction?

• Accuracy and Reliability: Our AI models are built to provide accurate data extraction, reduce errors, and improve efficiency.
• Scalability: Easily expand your document processing capabilities to meet your growing business needs.
• Customization: Tailor your extraction models to your specific needs, perfectly fitting your unique workflow.
• Grounded results: Localize data extracted from documents to generate responses from content, enabling human review workflows.
• Confidence Score: Leverage confidence scores to maximize the efficiency and minimize costs of your automated workflows.
• Cost-effectiveness: new priceExperience the world’s most advanced AI technology at a much lower price.

Extract custom fields model

The new field extraction model is available in: Azure AI Studio In AI services – Vision + document. First, create a project for working on your documents.

• Once you have selected your project, you should now be in the following location: Schema definition window. Your uploaded files will be listed and you can use the drop-down options to select them. Click to start adding fields. Add new field Click the button. Enter the name, description, and type of the fields you want to extract. Once all fields are added, select Save. Click the button at the bottom of the screen.

• After the schema is saved, all uploaded training documents are analyzed and field values ​​are automatically extracted. Automatically extracted fields are tagged. Predicted. Review the predicted values. If a field value is incorrect or not extracted, you can hover over the predicted field. Select the Edit button to make changes.After reviewing and correcting labels for all training documents, we proceed with model building.

• On the Build Model dialog page, provide a unique model name and optionally a description. SelectBuildStart the training process. The generative model will train immediately! When the status changes to Next, refresh the page to select the model. It was a success.
• Once model training is complete, you can test your model by selecting: Test button. Upload and select your test file Run Analysis Extract field values ​​from documents. Evaluate the results for each field to validate model accuracy.

You can use REST API or Client Library Submit your documents for analysis. Custom-generated AI models are very effective at extracting simple fields from documents without the need for labeled samples. However, providing a few labeled samples can significantly improve extraction accuracy for more complex fields and custom fields such as tables.

Business Scenario

• Loan and Mortgage Applications – Automating the loan and mortgage application process allows banks, lenders, and government agencies to process loan and mortgage applications more quickly.
• Financial Services – Analyze complex documents such as financial statements and asset management reports using new custom field extraction models.
• Contract Lifecycle Management – Build custom field extraction models to extract fields, clauses, and obligations from different types of contracts.
• Expense Management – You need to verify expenses by analyzing receipts and invoices from various retailers and businesses. Custom field extraction can extract expenses from various formats and documents using various templates.

Get started!

Custom generated models can be used with:2024-07-31-previewVersions and later models. Learn how to build and train. Custom field extraction model If you use generative AI, you can follow the instructions here. Build and train custom field extraction using AI Studio.. Start building your own custom document field extraction models today!