I started learning how to do machine learning on text data. I made a Python program with a small training set using the sklearn module to predict if a message is important or not. Since it is a small training set the results are not that accurate. This could be used for an app that learns what texts are considered important over time and marks them. Then the user manually decides if a text is actually important or not and the app learns what texts are considered important from this.
link to download the Python code
Here is an image of the output: