Is there a JSON data source for the recognition of emotions or a way to do one automatically?

I have a chatbot that tells us the weather when asked and I would like it to also recognize the emotions of the people who speak to motivate them when things are not going well. I want to start with the NLU and, therefore, provide you with data to understand the emotions of the people.

At the moment, I added lines in the data.json file to test if rasa_nlu is able to recognize emotions from them.

Here are some examples of data.json:

{
  "rasa_nlu_data": {
    "common_examples": [
      {
          ...
      },
      ...
      {
        "text": "I feel bad",
        "intent": "inform",
        "entities": [
          {
            "start": 7,
            "end": 10,
            "value": "bad",
            "entity": "sadness"
          }
        ]
      },
      {
        "text": "I feel happy",
        "intent": "inform",
        "entities": [
          {
            "start": 7,
            "end": 12,
            "value": "happy",
            "entity": "happiness"
          }
        ]
      }
    ]
  }
}

And here is nlu_model.py

Since writing scripts by hand is a long and tedious thing, my question is: how to generate more examples without having to add them by hand in the data.json file?

I am flexible about the types of feelings and if you have a database made, it suits me well too. For example in the folder /data of this project github I found a large csv file who could do the trick but this csv and I do not know if Rasa will accept it.