How to get 'language' field on the response to Cognitive Service-Read API call "Get Read Result" ?

Question

According to Cognitive Service Read API documentation, in particular documentation for the "Get Read Result" call get-read-result, the response should include a 'language' field.

When we perform our calls, we never get the 'language' field back.
We have tried specifying and not specifying the input language on the 'Read' call, but doesn't change the result.

Any hints or tips on why this is not working for us ?

Here an example of what we get:

{  
  "status": "succeeded",  
  "createdDateTime": "2021-06-10T10:16:07Z",  
  "lastUpdatedDateTime": "2021-06-10T10:16:07Z",  
  "analyzeResult": {  
    "version": "3.2.0",  
    "modelVersion": "2021-04-12",  
    "readResults": [  
      {  
        "page": 1,  
        "angle": 3.833,  
        "width": 2132,  
        "height": 1298,  
        "unit": "pixel",  
        "lines": [  
          {  
            "boundingBox": [  
              389,  
              434,  
              886,  
              417,  
              889,  
              578,  
              396,  
              592  
            ],  
            "text": "Hello",  
            "appearance": {  
              "style": {  
                "name": "other",  
                "confidence": 1  
              }  
            },  
            "words": [  
              {  
                "boundingBox": [  
                  407,  
                  441,  
                  887,  
                  418,  
                  888,  
                  581,  
                  413,  
                  592  
                ],  
                "text": "Hello",  
                "confidence": 0.76  
              }  
            ]  
          },  

.... cropped ....

        ]  
      }  
    ]  
  }  
}  
  

Answer 1

Hi, thanks for your feedback. The language detection feature is on our roadmap. By design, the read inference model actually does not detect language. You can leave language parameter as empty since read inference model supports multi-language, it should have the ability to detect most of languages by default. Sorry for any inconvenience, hope this helps!