This thesis presents recent work on a new method to automatically extract fine-grained duration information for common verbs using a large corpus of Twitter tweets. I present the results of distinguishing between habitual ...