This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(48, 90%, 14%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ELB%3C/text%3E%3C/svg%3E)
I trained a custom extraction model for invoices with around 45 labeled documents. However, none of the item values have a confidence score, even though the values are detected. This issue occurs in both the Form Recognizer Studio and the Json response. The "Amount" and "UnitPrice" values are of the subtype Currency, while the other values are strings. I noticed that when using the same document with the pre-built invoice model, the values were extracted correctly. Can someone help me understand what might be causing the problem with my custom model?
Screenshot of the issue:
Screenshot of the pre-built model extraction:
The api-version of the custom model is "2023-07-31"
*Edit: The same problem appears at the "TaxDetails" in receipts.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(6.4, 61%, 10%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ES%3C/text%3E%3C/svg%3E)
Hi @Leonhard Blank ,
Thank you for reaching out to Microsoft Q&A forum!
As per your query, the values are extracted correctly with the pre-built invoice model, but there is an issue with your custom model. Though you have trained a custom extraction model for invoices with around 45 labeled documents. However, none of the item values have a confidence score, even though the values are detected.
The possible reason could be that the custom model is not trained with enough variation in the data. If the training data is too similar, the model may not be able to differentiate between different types of invoices. I suggest you to train the data has enough variation to cover different types of invoices. Another thing you can try is to analyze the JSON response from the custom model to see if there are any errors or warnings that may indicate the cause of the issue.
The confidence scores based on the probability that the prediction is correct. Document Intelligence analysis results return an estimated confidence for predicted words, key-value pairs, selection marks, regions, and signatures. Currently, not all document fields return a confidence score.
I hope this information helps! Thank you.
Hi @santoshkc , thanks for the reply!
So actually I cold not find any errors in the JSON response.
Also I think the balance isn’t causing this problem. Every other field than Items has a high confidence score.
Greetings!
Hi @Leonhard Blank ,
Thank you for your response.
Currently, not all document fields return a confidence score and Custom neural models do not provide accuracy scores during training.
Do let us know if you have any further questions. Thank you.
According to documentation, there is no way to get the confidence scores for structured data (like the Items table) when using custom models.