wiki.ziemers.de

Microsoft Azure is a Cloud computing platform from Microsoft. Our primary used service is the cognitive services text analytics. For the Beuthbot we use the student version off Azure.We chose this provider because we did not need to provide a credit card here.

1. Microsoft Azure
2. Table of content
3. Student Account
4. Cognitive Services Text Analyticst
5. Knowledge

The Student Account is an free to use Account from Microsoft Azure. Here we get some limittet free acess to different services of the azure programm. Details you can find under: https://azure.microsoft.com/de-de/free/students/

Because we have problems when creating the account, we have designed a following short manual:

1. click on „activate now“ on https://azure.microsoft.com/de-de/free/students/
2. login with an privat microsoft account(no University Mail!)
3. after login you must verify your Student Account with University Mail
4. at last you must ident you by name, mail address and phone number here you get an activation code what you must

The important knowledge what we get from this registration is that we need private microsoft accounts and that the microsoft support is not very helpful.

To use the ervice we create in oure Azure account an BeuthBot Projekt(resourcesgroup). In this we create an Cognitive Services Text Analytics in North Europe. Now we get an Api End Point from Microsoft where we can do „Post „request with our messages. We have four options to ask for: the language, analyze sentiment, Extract key phrases and Identify linked entities.

More informations about the API you can find her: https://docs.microsoft.com/de-de/azure/cognitive-services/text-analytics/.

For the moment we use the service direktly in our gateway later we want to outsource the service in an extra NLU Request Service(Adapter for NLU).

A Post Request Object in JavaScript should look like the following example:

POST https://beutbot.cognitiveservices.azure.com/text/analytics/v3.0-preview.1/languages HTTP/1.1
 
Host: beutbot.cognitiveservices.azure.com
 
Content-Type: application/json 
 
Ocp-Apim-Subscription-Key: •••••••••••••••••••••••••••••••• 
 
 
{ 
  "documents": [
    {
      "id": "",
      "text": "" 
    }
  ] 
}

The response body has following structure:

{
   "documents": [
      {
         "id": "",
         "detectedLanguages": [
            {
               "name": "",
               "iso6391Name": "",
               "score": 
            }
         ]
        }
    ]
 }

To interpret the response we have to write a parser for all operations what we ask by the API.

Very quickly, we realized that is it not possible to operate a chatbot with the text analytics services without paying for the service. Because the free limited acess of the service is consumed very quickly in our own tests. So, while we can show that it is possible to use a chatbot with this service, it will not be usable that way. We allready found out, that IBM has an also free NLU Service with more request than azure but there we will have the same problem but we try to implement this service also, so that we cann show that our NLU Adapter work and we have different options for services to request.