How do I extract tables from azure through C# API

Question

How do I extract tables from azure through C# API

Vijay Prime 0

User's image

Hello,
Can you please tell me how to extract tables from Azure through C# API. I am trying to get specific details from the MTC documents and for that I have created and trained a custom neural model for it and it is mapping the values of the table correctly.

The issue: Tables are returning as an empty array when I call the model through the API.

The single key value pair fields are extracting correctly when I use the API but the tables are returning as empty arrays. How do I extract the tables?
Do you have a sample code for table extraction?

SriLakshmi C 4,875 Reputation points Microsoft External Staff Moderator

2025-05-09T16:10:45.6566667+00:00
Hello @Vijay Prime,

To extract tables from Azure using the C# API, you can utilize the AnalyzeResult.Tables property from the Azure Document Intelligence SDK. This property provides a read-only list of extracted tables from the analyzed document.

Make sure your custom neural model is adequately trained to recognize and extract tables from the documents. If the model hasn’t fully learned the structure or patterns of tables during training, it may result in empty or incomplete outputs.

When invoking the model via the API, be sure to review the full response carefully. It may contain additional context, warnings, or error messages that can assist in diagnosing and resolving the issue.

Check if the input document type is supported for table extraction (e.g., table extraction is not supported for XLSX files).

If your tables span multiple pages, consider splitting the document into individual pages before analysis.

Here is a simple example of how you can extract tables using the Azure.AI.DocumentIntelligence library in C#:

using Azure; using Azure.AI.DocumentIntelligence; using Azure.AI.DocumentIntelligence.Models; // Instantiate the client var client = new DocumentAnalysisClient(new Uri("your-endpoint"), new AzureKeyCredential("your-key")); // Call the AnalyzeDocument method var operation = await client.StartAnalyzeDocumentAsync("prebuilt-document", yourDocumentStream); var response = await operation.WaitForCompletionAsync(); // Access the tables from the response var tables = response.Value.AnalyzeResult.Tables; if (tables.Count > 0) { foreach (var table in tables) { Console.WriteLine($"Table found with {table.RowCount} rows and {table.ColumnCount} columns."); } } else {

Make sure to replace your-endpoint, your-key, and yourDocumentStream with the appropriate values.

Try running the API call on a simpler or different document to ensure that the extraction works under various conditions.

I Hope this helps. Do let me know if you have any further queries.

Thank you!

1 answer

Your answer

Answer 1

Hi @Vijay Prime

Thank you for using Azure Document Intelligence. Based on screenshot (thanks for providing), you've trained a custom model, and the "Mechanical Properties" field is of type 'Fixed Table'.

In the JSON API response, you’ll find this table under (example for 'Tests''Hardness' cell):

analyzeResult.documents[0].fields["Mechanical Properties"].valueObject["Tests"].valueObject["Hardness"]

'Fixed Table' fields are represented as a "dictionary of dictionary".

Below is a C# sample (based on Custom_extraction_model.cs ) showing how to process 'Fixed Table' output:

//...
DocumentIntelligenceClient client = new DocumentIntelligenceClient(new Uri(endpoint), credential);
var analyzeOptions = new AnalyzeDocumentOptions(modelId, BinaryData.FromBytes(File.ReadAllBytes("c:\\temp\\SampleMechProp.png"))); // use your own local file
Operation<AnalyzeResult> operation = await client.AnalyzeDocumentAsync(WaitUntil.Completed, analyzeOptions);
AnalyzeResult result = operation.Value;
Console.WriteLine($"Document was analyzed with model with ID: {result.ModelId}");
foreach (AnalyzedDocument document in result.Documents)
{
    foreach (var (fieldName, field) in document.Fields)
    {
        Console.WriteLine($"Field '{fieldName}': ");

        // Illustration how to read output of Fixed Tables.
        if (field.FieldType == DocumentFieldType.Dictionary) 
        {
            var columnHeaders = field.ValueDictionary.SelectMany(rowField => rowField.Value.ValueDictionary.Keys).Distinct().ToList();
            Console.WriteLine("{0,-15} {1}", "_RowKey_", string.Join(" ", columnHeaders.Select(h => $"{h,-15}")));
            Console.WriteLine(new string('-', 15 + columnHeaders.Count * 15));
            foreach (var (rowKey, rowValue) in field.ValueDictionary)
            {
                // Using value.Content for simplicity, but value.Value%Type% can be used to get normalized value of the subfield.
                var rowValues = columnHeaders.Select(subFieldName => rowValue.ValueDictionary.TryGetValue(subFieldName, out var value) ? value.Content ?? "" : "").ToList(); 
                Console.WriteLine("{0,-15} {1}", rowKey, string.Join(" ", rowValues.Select(v => $"{v,-15}")));
            }
        }
        else
        {
            Console.WriteLine($"Content: '{field.Content}'");
        }
    }
}

I trained a sample model (similar to your) and ran this code. It prints:

Document was analyzed with model with ID: SampleFixedTableCustomer
Field 'Mechanical Properties':
_row_           Hardness        Elongation %
---------------------------------------------
Min             137             22.00
Max             197
Actual          143,146,149     30.46

Please let me know if this doesn’t work for you.

Note: Install and reference the Azure.AI.DocumentIntelligence C# SDK v1.0.0 (not any 1.0.0-beta.x versions).

Best Regards,
=Anatoly=

Share via

How do I extract tables from azure through C# API

1 answer

Your answer