Error in Azure AI Search with OCR Skill

Question

Hi, I was creating an AI Search index to on my png files. I followed the instruction from the below sources, but still getting errors. Please help me out on where I missed in the code.

• https://learn.microsoft.com/en-us/azure/search/cognitive-search-skill-image-analysis
• https://learn.microsoft.com/en-us/azure/search/cognitive-search-skill-ocr

I got the following errors when running the indexer.

The data field 'myLayoutText' in the document with key 'xxxxx' has an invalid value of type 'Collection(Edm.ComplexType)' ('JSON arrays with element type 'Object' map to Collection(Edm.ComplexType)'). The expected type was 'Edm.ComplexType'.

Here is my code:

headers = {
    'Content-Type':'application/json',
    'api-key':search_key
}

data_source_name = 'document-bot-files-datasource'


index_name = 'ocr-index'
index_data = {
        "name" : index_name,
        "fields": [
            { "name": "ID", "type": "Edm.String", "key": True, "searchable": True,"filterable": True, "sortable": True,"facetable":True,"analyzer":"keyword" },
            { "name": "parent_id", "type": "Edm.String", "searchable": True, "filterable": True, "sortable": False,"facetable":False},
        
            { "name": "myText", "type": "Edm.String", "searchable": True, "filterable": False, "sortable": False,"facetable":False},
            { "name": "myLayoutText", "type": "Edm.ComplexType","fields":[
                { "name": "language", "type": "Edm.String", "searchable": False, "filterable": False, "sortable": False,"facetable":False},
                { "name": "text", "type": "Edm.String", "searchable": True, "filterable": False, "sortable": False,"facetable":False},
                { "name": "lines", "type": "Collection(Edm.ComplexType)","fields":[
                    { "name": "boundingBox", "type": "Collection(Edm.ComplexType)", "fields":[
                        { "name": "x", "type": "Edm.Int32", "searchable": False, "filterable": False, "sortable": False,"facetable":False},
                        { "name": "y", "type": "Edm.Int32", "searchable": False, "filterable": False, "sortable": False,"facetable":False},
                    ]},
                    { "name": "text", "type": "Edm.String", "searchable": True, "filterable": False, "sortable": False,"facetable":False},
                ]},
                { "name": "words", "type": "Collection(Edm.ComplexType)","fields":[
                    { "name": "boundingBox", "type": "Collection(Edm.ComplexType)", "fields":[
                        { "name": "x", "type": "Edm.Int32", "searchable": False, "filterable": False, "sortable": False,"facetable":False},
                        { "name": "y", "type": "Edm.Int32", "searchable": False, "filterable": False, "sortable": False,"facetable":False},
                    ]},                    
                    { "name": "text", "type": "Edm.String", "searchable": True, "filterable": False, "sortable": False,"facetable":False},
                ]}
            ]},
        ],
        "vectorSearch": {
        "profiles": [
        {
            "name": "myHnswProfile",
            "algorithm": "myHnsw"
        }
        ],
        "algorithms": [
        {
            "name": "myHnsw",
            "kind": "hnsw",
            "hnswParameters": {
            "m": 4,
            "metric": "cosine"
            }
        }
        ]
    }
    }
index_delete_response = requests.delete(f"{search_endpoint}/indexes/{index_name}?api-version=2020-06-30",headers=headers)
index_response = requests.put(f"{search_endpoint}/indexes('{index_name}')?api-version=2023-11-01",json=index_data,headers=headers)
print("index creation:", index_response.text)

skill_name = 'ocr-skillset'
skill_data = {
  "skills": [
    {
      "description": "Extracts text (plain and structured) from image.",
      "@odata.type": "#Microsoft.Skills.Vision.OcrSkill",
      "context": "/document/normalized_images/*",
      "defaultLanguageCode": '',
      "detectOrientation": True,
      "inputs": [
        {
          "name": "image",
          "source": "/document/normalized_images/*"
        }
      ],
      "outputs": [
        {
          "name": "text",
          "targetName": "myText"
        },
        {
          "name": "layoutText",
          "targetName": "myLayoutText"
        }
      ],
    }
  ]
}
skill_update_response = requests.put(f"{search_endpoint}/skillsets('{skill_name}')?api-version=2023-10-01-Preview",json=skill_data,headers=headers)
print("skill creation update code:", skill_update_response.status_code)
# print("skill creation update code:", skill_update_response.text)

indexer_name = 'ocr-indexer'
indexer_data = {
    "name" : indexer_name,
    "dataSourceName" : data_source_name,
    "targetIndexName" : index_name,
    "skillsetName":skill_name,
    "parameters": {
        "configuration": {
            "indexedFileNameExtensions" : ".pdf,.docx,.txt,.png,.jpeg",
            "dataToExtract": "contentAndMetadata",
            "parsingMode": "default",
            "imageAction": "generateNormalizedImages"
        }
    },
    "schedule" : { },
    "fieldMappings" : [
    ],
    "outputFieldMappings": [
        {
            "sourceFieldName": "/document/normalized_images/*/myText",
            "targetFieldName": "myText"
        },
        {
            "sourceFieldName": "/document/normalized_images/*/myLayoutText",
            "targetFieldName": "myLayoutText"
        },
    ]
    }
indexer_update_response = requests.put(f"{search_endpoint}/indexers/{indexer_name}?api-version=2020-06-30",json=indexer_data,headers=headers)
print("indexer update  code: ",indexer_update_response.status_code)
indexer_reset_response = requests.post(f"{search_endpoint}/indexers/{indexer_name}/reset?api-version=2020-06-30",headers=headers)
print("indexer reset code: ",indexer_reset_response.status_code)
# time.sleep(3)
indexer_run_response = requests.post(f"{search_endpoint}/indexers/{indexer_name}/run?api-version=2020-06-30",headers=headers)
print("indexer rerun code: ",indexer_run_response.status_code)

Answer 1

Hello Violet !

Thank you for posting on Microsoft Learn Q&A.

The OCR layoutText output is emitted per image but your index schema declares myLayoutText as a single Edm.ComplexType. In practice it is recommended storing both OCR outputs (text and layoutText) as Collection(Edm.String) which means each element is a JSON string for one normalized image.

https://learn.microsoft.com/en-us/azure/search/cognitive-search-concept-image-scenarios

You need to change the index fields to string collections:

index_data = {
  "name": "ocr-index",
  "fields": [
    { "name": "ID", "type": "Edm.String", "key": true, "searchable": true, "filterable": true, "sortable": true, "facetable": true, "analyzer": "keyword" },
    { "name": "parent_id", "type": "Edm.String", "searchable": true, "filterable": true },
    
    { "name": "myText",       "type": "Collection(Edm.String)", "searchable": true },
    { "name": "myLayoutText", "type": "Collection(Edm.String)", "searchable": true }
  ],
  "vectorSearch": { ... }
}