How to predict tags in Stack Overflow questions using Keras

Through the public Stack Overflow dataset, I will predict the topic/label for each given Stack Overflow question. There are 20 labels such as c#, python, css, javascript, androind, iphone, ios, etc, and there are 40.000 questions. I will split the data into train and test. Using Keras, a python deep learning library, I will process the text from both questions and labels. Once the text has been converted to a format Keras can read, I will apply the model. I will then make predictions on test data. I can also ask audience to come up with new questions and ask the model what it predicts.