Auspice is an app designed to predict how a particular Twitter user will react to a specific tweet from a specific account. Auspice is particularly useful for predicting how users will respond to promoted tweets.
Auspice was trained using gradient boosting on a set of approximately 10,000 responses to promoted tweets. These tweets were classified by sentiment using the VADER sentiment analyzer. The classifier uses a set of features about the account publishing the initial tweet, the initial tweet itself, and the account whose response is being prediction. Many of these features were derived using GloVe.
According to Dictionary.com, the word "auspice" is derived from the Latin "auspicium," meaning "divination from the flight of birds."