Using AI Model Hub (Pre-Trained Document Types)

Docsumo's Pre-Trained Document Types offer a seamless way to extract data from your documents. This guide will walk you through:

  1. Using a document type from AI models hub and reviewing your documents.
  2. Modifying extracted fields
  3. Training an advanced model for the new fields.

Using a Document Type from AI Models Hub and Reviewing your Documents.

Adding a pre trained document type from AI Models Hub and utilising it's default configuration.

Step 1. Accessing Pre-Trained Document Types

  • Log in to your Docsumo account.
  • Click on "AI Models Hub" in the left navigation menu or please click on add new document types in my document page.

Step 2. Choosing a Pre-Trained Document Type

  • Browse the available document types categorized in the AI Models Hub.
  • Find the document type that matches the documents you need to process.

Step 3. Adding the Document Type

  • Click "Use" on the desired document type to access its details.
  • Name your document type and upload the initial documents. You can upload 1 or more documents basis your convenience. Alternatively you can also use the sample document option instead of uploading any of your documents by clicking "Proceed with sample document type name".
  • Click "Create" to start uploading the document. You do not have to click "Create" if you choose to proceed with the Sample document option.

Step 4. Review the Document for Extraction

  • Once the document is processed and ready to use, click "Review" from the upload widget. You can alternatively navigate to the review screen from the Document type card(click the "Review" button) and My Document page(click on the name of the document that you uploaded) as well.
  • Check the document for the fields and the data that has been extracted.
  • Click "Confirm" if you find the extraction correct. You can also download the final output from here in the form of Excel or JSON file.

Modifying Fields

Based on your requirement, you might be needing some unique fields or update any existing fields names for the data that is being extracted.

Step 1. Add or Update Fields as per your Requirement.

Step 2. Upload New Documents or Review an Old Document to See the Changes.

  • Based on your selection to applying the changes of updating the fields, you can view the changes on the review screen over new or old documents as well.
  • While you might get to see the new fields or updated names, extraction might happen around them. This is happening as the Model is not trained for these new fields.

Training an Advanced Model for the New Fields

If you need to extract new fields not covered by the pre-trained model, it is important to train a advanced model for the new fields to ensure automated extraction of data.

Step 1. Annotate 20 Documents

  • Annotate at least 20 documents with the new fields and confirm these documents. AI Assist will help you at this by providing suggestions to expedite annotation.
  • Before confirming the documents, ensure that all the fields extracted are correctly annotated or extracted.

Step 2. Training the Model

  • After annotating sufficient documents, initiate the model training process. Follow the simple steps here to train an advanced model.

Step 3. Checking Model Accuracy

  • Once a model is trained and linked to the document type, upload new documents to see how the system performs with the modified and trained model.
  • You can can check the accuracy and retrain model after your have approved 50+ document using step 2.

By following this guide, you can effectively utilise Docsumo's Pre-Trained Document Types, customize extracted fields, train new fields, and enhance your document processing capabilities to meet your specific needs.