More Features Are Not Always Better: Evaluating Generalizing Models in Incident Type Classification of Tweets

34848

Knowledge Triples