We only build models to predict if the tweet will be liked or reshared or commented. What I am so confused about is how we would rank them. It seems like we need to use the engagement scores for each tweet. However, how should we calculate the engagement scores? It seems different from the user engagement scores mentioned in the early sessions.