Improve accuracy

[chaitanya928]

How can I improve the accuracy of data extraction in document understanding using the extractors prebuilt existing

[AutomationX_by_Kiran]

Hi @chaitanya928 ,

There are several ways to improve the accuracy of data extraction in Document Understanding:

1. Use a high-quality and relevant training dataset: Make sure your training dataset includes a diverse set of documents that are representative of the documents you want to process. The more representative your training data, the better your model will perform on new documents.
2. Fine-tune the extraction parameters: Experiment with different configurations and parameters for the extractor, such as the confidence threshold, to find the best settings for your use case.
3. Use a custom extractor: If the predefined extractors are not performing well on your documents, consider training a custom extractor using the AI Center. This will allow you to tailor the extraction process to your specific needs.
4. Leverage human-in-the-loop (HITL) processes: If your documents are complex or ambiguous, consider implementing a human-in-the-loop process where human reviewers can validate and correct the extracted data. This will help improve the overall accuracy of the extraction process.

Regards
Happy Automation !!!

[Dilli_Reddy]

@chaitanya928

Follow this Steps:

Gather a diverse set of high-quality training documents that represent the variability in your documents. The quality and diversity of your training data play a crucial role in model accuracy.
Annotate or label your training data with the expected data extraction results. This helps the model learn what data to extract.
Train your model iteratively, analyzing the results after each training cycle.
Use the provided fine-tuning options in Document Understanding to adjust the prebuilt extractors to your specific document types and layouts.
Maintain version control of your trained models. You can revert to a previous version if a newer one performs worse, ensuring continuous reliability.
Always test your extraction models on real, representative documents from your workflow to ensure they work as expected.

Cheers…!