locked
Help needed for deploying a statistical model for text mining developed with R RRS feed

  • Question

  • Hello,

    I have developed a statistical model that actually input keywords, find it in sentences that contain it, and further classify in emotion classes. It is a lexicon based emotion classification takes input from three modules (keywords, sql database, emotion related word dictionary) and gives output in 10 different modules divided per month. Now I need to publish it as webservice (I have already read documents how to publish as web service). But for me I cannot get any option to specify my input schema and testing the webservice also not working. I developed the whole thing with R and its different packages. My model is something like this, 

    Sunday, May 22, 2016 4:23 PM

Answers

  • Please see document which explains the concepts and how to use them:

    https://azure.microsoft.com/en-us/documentation/articles/machine-learning-publish-a-machine-learning-web-service/

    Thanks,

    Raymond

    • Marked as answer by CharisLove Tuesday, May 24, 2016 12:43 AM
    Monday, May 23, 2016 5:21 PM

All replies

  • Input schema is derived from the port where web service input port is setup. So you will need to setup multiple input and multiple output ports and use Excel to test it

    Can you please share your  web service URL?

    Sunday, May 22, 2016 7:49 PM
  • surely I can share the URL with you. But it is not published yet. And I did not understand your point, when I am going to publish as web service, it is not giving me option to define the schema. Can you suggest any documentation?
    Monday, May 23, 2016 2:36 AM
  • You don't explicitly define a input schema when you publish a web service. Rather, you help system infer the schema by adding a web service input port into the graph, connect to a port of the module where your input data will actually be fed to. Meanwhile, there should be another module also connects to that same port. The schema of that incoming edge will become the schema of your web service input. Some times it can be ambiguous and leads to error, especially when you use Enter Data Manually module. In that case, it is best to also connect it to one or more Edit Metadata module to make the input types explicit. This will help the inference job.
    Monday, May 23, 2016 3:13 AM
  • Thanks for your great tip, especially about enter data manually. I am trying to follow as you advised.
    Monday, May 23, 2016 5:24 AM
  • It is not working for me,  I cannot see any  "Create Scoring Experiment" option in Azure Studio from where I can define the input outputs.

    Monday, May 23, 2016 3:30 PM
  • Please see document which explains the concepts and how to use them:

    https://azure.microsoft.com/en-us/documentation/articles/machine-learning-publish-a-machine-learning-web-service/

    Thanks,

    Raymond

    • Marked as answer by CharisLove Tuesday, May 24, 2016 12:43 AM
    Monday, May 23, 2016 5:21 PM
  • Thanks Raymond, 

    I have already read the document, but it could not solve my problem. For a machine learning problem there is "Scoring Module", so after running the project, it gives "create scoring experiment" button, that leads to "deploy web service". Or there is another button "set up web service" available for machine learning project that leads to "deploy web service". For me, none of these two buttons are present in my work space, what I get only "Deploy Web Service" button, that definitely not helping defining the inputs and outputs, I tried so but in test model there is nothing. here is the screen shot:

    Tuesday, May 24, 2016 5:46 AM