Extract line item fields with query fields

Question

Extract line item fields with query fields

Maher 21

I've seen a few posts that says its possible to extract line item fields with query fields. I've tried this and at best it will only extract a single field value and not extract the value for every line.

If there is a way to extract additional line item fields with query fields for each line please let me know how ideally with an example.

Thanks!

Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-15T15:44:24.08+00:00

Hi Maher

I am able to find multiple line-item fields using "prebuilt Receipt" models but filtering with query fields only gives the first row.

I have requested sample data in private chat to replicate and suggest best model as per.

Thank you
Maher 21 Reputation points

2025-04-15T15:51:34.5766667+00:00

Ideally I'm trying to do this programmatically. I know custom extraction works but I don't wnat to have to rely on it so much.
Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-15T15:59:27.4533333+00:00

Hi Maher

Yes, mentioning query field only fetched on field from 1st line.

But we can extract all the field values programmatically using SDK without mentioning query field.

Attached sample code on prebuilt receipt model for reference.

Thank you.
Maher 21 Reputation points

2025-04-15T16:02:01.8133333+00:00

Your code only has standard fields. I'm trying to extract non-standard fields from the line items.
Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-15T16:06:25.45+00:00

Could you share sample files and mention the non-standard fields in private chat or with screenshot here in forum.

It would help a lot to replicate and suggest further.

Thank you.

Manas Mohanty 5,700 Microsoft External Staff Moderator

Hi Maher

I understood your point on extracting non-standard (not numeric, alphanumeric or Data etc.). Shall repro with drafted data based on below schema and update you.

{
  "StandardID": "STD78901",  // Standard field
  "RequirementID": "REQ45678",  // Standard field
  "CreatedDate": "2025-04-16",  // Standard field
  "ModifiedDate": "2025-04-16",  // Standard field
  "RequirementTitle": "Pilot Pathway Choice",  // Non-standard field
  "PilotStartDate": "2025-05-01",  // Non-standard field
  "Owner": "StandardsLedgerUser",  // Non-standard field
  "Overview": "Overview of the requirement",  // Non-standard field
  "SecurityReviewDecision": "Approved",  // Non-standard field
  "SemesterPriority": "High",  // Non-standard field
  "AdoptionEffort": "Medium",  // Non-standard field
  "CohortDefinition": "Group A",  // Non-standard field
  "ExistingTelemetry": "Telemetry data",  // Non-standard field
  "YearAndMonthCreated": "2025-04"  // Standard field
}

Thank you.

Manas Mohanty 5,700 Microsoft External Staff Moderator

Hello Maher

I think you can use the OCR/Read model to filter out non-standard values.


result = poller.result()

for page in result.pages:
        print(f"----Analyzing document from page #{page.page_number}----")
        print(f"Page has width: {page.width} and height: {page.height}, measured with unit: {page.unit}")

        # Analyze lines
        if page.lines:
            for line_idx, line in enumerate(page.lines):
                words = get_words(page, line)
                print(
                    f"...Line # {line_idx} has {len(words)} words and text '{line.content}' within bounding polygon '{line.polygon}'"
                )

                # Analyze words
                for word in words:
                    print(f"......Word '{word.content}' has a confidence of {word.confidence}")
        
        # Extract non-standard fields
        for field in page.fields:
            if field.name not in ["InvoiceNumber", "Date", "CustomerName", "Quantity", "UnitPrice"]:
                non_standard_fields[field.name] = field.value
    
            print("Non-standard fields:", non_standard_fields)

Please let us know if OCR model suits your requirement.

Thank you

Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-17T06:01:36.61+00:00

Hello Maher

We have not heard from you. Hope the pointers shared were useful to you.

Thank you.
JAYA SHANKAR G S 4,035 Reputation points Microsoft External Staff Moderator

2025-04-25T08:48:32.3166667+00:00

Hello @Maher ,

As per this documentation when you want to extend the schema apart from what you have then you can go with query fields.

And Query fields return a single answer (or best match) for each query like Date fields, they’re not intended to iterate over line items like a table.

So, if you line items can't be described as a key value pair in the document then you may need to go with custom model and if you could provide a sample pdf which is similar to which you trying it will be helpful to reproduce the scenario and check best model for it from our end

Thank you
JAYA SHANKAR G S 4,035 Reputation points Microsoft External Staff Moderator

2025-04-28T03:43:34.7766667+00:00

Hello @Maher ,

Please check above suggestion on query fields and provide your sample pdf file which helps to reproduce the scenario and check best model for it from our end.

Thank you

1 answer

Your answer

Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-15T15:44:24.08+00:00

Hi Maher

I am able to find multiple line-item fields using "prebuilt Receipt" models but filtering with query fields only gives the first row.

I have requested sample data in private chat to replicate and suggest best model as per.

Thank you
Maher 21 Reputation points

2025-04-15T15:51:34.5766667+00:00

Ideally I'm trying to do this programmatically. I know custom extraction works but I don't wnat to have to rely on it so much.
Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-15T15:59:27.4533333+00:00

Hi Maher

Yes, mentioning query field only fetched on field from 1st line.

But we can extract all the field values programmatically using SDK without mentioning query field.

Attached sample code on prebuilt receipt model for reference.

Thank you.
Maher 21 Reputation points

2025-04-15T16:02:01.8133333+00:00

Your code only has standard fields. I'm trying to extract non-standard fields from the line items.
Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-15T16:06:25.45+00:00

Could you share sample files and mention the non-standard fields in private chat or with screenshot here in forum.

It would help a lot to replicate and suggest further.

Thank you.
Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-16T08:45:15.06+00:00

Hi Maher

I understood your point on extracting non-standard (not numeric, alphanumeric or Data etc.). Shall repro with drafted data based on below schema and update you.

{ "StandardID": "STD78901", // Standard field "RequirementID": "REQ45678", // Standard field "CreatedDate": "2025-04-16", // Standard field "ModifiedDate": "2025-04-16", // Standard field "RequirementTitle": "Pilot Pathway Choice", // Non-standard field "PilotStartDate": "2025-05-01", // Non-standard field "Owner": "StandardsLedgerUser", // Non-standard field "Overview": "Overview of the requirement", // Non-standard field "SecurityReviewDecision": "Approved", // Non-standard field "SemesterPriority": "High", // Non-standard field "AdoptionEffort": "Medium", // Non-standard field "CohortDefinition": "Group A", // Non-standard field "ExistingTelemetry": "Telemetry data", // Non-standard field "YearAndMonthCreated": "2025-04" // Standard field }

Thank you.
Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-16T11:06:26.7266667+00:00

Hello Maher

I think you can use the OCR/Read model to filter out non-standard values.

result = poller.result() for page in result.pages: print(f"----Analyzing document from page #{page.page_number}----") print(f"Page has width: {page.width} and height: {page.height}, measured with unit: {page.unit}") # Analyze lines if page.lines: for line_idx, line in enumerate(page.lines): words = get_words(page, line) print( f"...Line # {line_idx} has {len(words)} words and text '{line.content}' within bounding polygon '{line.polygon}'" ) # Analyze words for word in words: print(f"......Word '{word.content}' has a confidence of {word.confidence}") # Extract non-standard fields for field in page.fields: if field.name not in ["InvoiceNumber", "Date", "CustomerName", "Quantity", "UnitPrice"]: non_standard_fields[field.name] = field.value print("Non-standard fields:", non_standard_fields)

Please let us know if OCR model suits your requirement.

Thank you
Manas Mohanty 5,700 Reputation points Microsoft External Staff Moderator

2025-04-17T06:01:36.61+00:00

Hello Maher

We have not heard from you. Hope the pointers shared were useful to you.

Thank you.
JAYA SHANKAR G S 4,035 Reputation points Microsoft External Staff Moderator

2025-04-25T08:48:32.3166667+00:00

Hello @Maher ,

As per this documentation when you want to extend the schema apart from what you have then you can go with query fields.

And Query fields return a single answer (or best match) for each query like Date fields, they’re not intended to iterate over line items like a table.

So, if you line items can't be described as a key value pair in the document then you may need to go with custom model and if you could provide a sample pdf which is similar to which you trying it will be helpful to reproduce the scenario and check best model for it from our end

Thank you
JAYA SHANKAR G S 4,035 Reputation points Microsoft External Staff Moderator

2025-04-28T03:43:34.7766667+00:00

Hello @Maher ,

Please check above suggestion on query fields and provide your sample pdf file which helps to reproduce the scenario and check best model for it from our end.

Thank you

Answer 1

Hey Maher ,

Thanks for the question on Microsoft Q&A!

I got that query fields aren’t pulling all line items for you ,just one value. They’re not great for tables. For invoices, use the prebuilt-invoice model; it grabs all line items like item name, quantity, and price automatically. For other docs, try the prebuilt-layout model to get table data, then sort it out.

Example: A table in a receipt might have “Item,” “Qty,” “Cost” as headers, with rows like “Bread, 2, $4.00” and “Milk, 1, $3.00.” You’d make those into {"Item": "Bread", "Qty": "2", "Cost": "4.00"} for each.

Hope this helps! If it works, please upvote and mark as accepted. Thanks!

References:

Share via

Extract line item fields with query fields

1 answer

Your answer