Topicmodeling for large-scale text data