How to increase the accuracy on the pre-built invoice model or general layout model on Document Intelligence Studio?

Question

The pre-built invoice model misinterpreted the invoice I provided. It incorrectly identified the vendor name and address, merging the vendor name into the vendor address field, which is inaccurate. Is there a way to provide personalized input (e.g., prompt engineering) to improve the model's accuracy?

Answer 1

Hi @Jessie Chen,

Thank you for reaching out to Microsoft Q&A forum!

To increase the accuracy of the pre-built invoice model or general layout model in Azure Document Intelligence Studio, consider the following approaches:

1. Increase the number of Documents: Provide a larger volume of training documents to the model. A greater quantity of data allows the model to learn from more examples, which can lead to improved accuracy.
2. Diversify the Document Types: Ensure that your training data includes a wide variety of invoice formats and layouts. This diversity helps the model generalize better across different styles and structures, reducing the likelihood of misinterpretations.
3. Use Custom Models: If the pre-built model continues to misinterpret specific fields, consider creating a custom extraction model tailored to your unique invoice formats. Custom models can be trained with your own labeled data, improving accuracy on documents specific to your use case.

It is important to note that providing personalized input (e.g., prompt engineering) to improve the model's accuracy is not currently supported in the Azure Document Intelligence framework.

By employing above strategies, you can significantly improve the accuracy of the Azure Document Intelligence models for invoice or any document processing.Hope this helps. Do let us know if you any further queries.

---

If this answers your query, do click Accept Answer and Yes for was this answer helpful.