We consider the problem of recognizing mentions of human senses in text. Our contribution is a method for acquiring labeled data, and a learning method that is trained on this data. Experiments show the effectiveness of our proposed data labeling approach and our learning model on the task of sense recognition in text.