Note that the first two steps were automatically done for you if you used the Deploy to Bluemix button. You still need to train George, though.

$ cf create-service natural_language_classifier free george-classifier

Assuming you want to call your app my-own-george:

$ cf push my-own-george

This will make your app available at http://my-own-george.mybluemix.net (or http://my-own-george.eu-gb.mybluemix.net if you're using the UK region).

But until the classifier is trained, George won't be able to answer any questions.

First you need to retrieve the credentials for the classifier service you created above. You can find them on the Bluemix console if you look hard enough, or you can find them by dumping the application's environment with this command (substituting my-own-george with your app name, of course):

$ cf env my-own-george
…
{
 "VCAP_SERVICES": {
  "natural_language_classifier": [
   {
    "credentials": {
     "password": "ppppppppppp",
     "url": "https://gateway.watsonplatform.net/natural-language-classifier-experimental/api",
     "username": "uuuuuuuu-uuuu-uuuu-uuuu-uuuuuuuuuuuu"
    },
…

Armed with these username and password, you can now create a classifier instance, by using the corpus.json provided with George:

$ curl -Ss -u uuuuuuuu-uuuu-uuuu-uuuu-uuuuuuuuuuuu:ppppppppppp \
    -H "Content-Type:application/json" \
    -d @corpus.json \
    https://gateway.watsonplatform.net/natural-language-classifier/api/v1/classifiers
{
  "classifier_id": "ABCDEF-nlc-123",
  "name": null,
  "created": "2015-06-24T12:56:58.555Z",
  "url": "https://gateway.watsonplatform.net/natural-language-classifier/api/v1/classifiers/ABCDEF-nlc-123",
  "status": "Training",
  "status_description": "The classifier instance is in its training phase, not yet ready to accept classify requests"
}

As you can see, the classifier is training itself. It will stay like this for a few minutes.

However, now that we have a classifier ID (ABCDEF-nlc-123 above), we don't have to wait to configure our application to use it:

$ cf set-env my-own-george CLASSIFIER_ID ABCDEF-nlc-123
$ cf restage my-own-george

Once the classifier is ready (which you can easily check by accessing its full URL as returned above), George should be ready to answer questions.

Anytime you change corpus.json, you will have to train a new classifier instance, and update your app's CLASSIFIER_ID environment variable accordingly.

See here for the full Natural Language Classifier API reference.