Broadcast News is a large text dataset for next word prediction that includes the 1996 Broadcast News Speech Corpus, which contains 104 hours of broadcasts from the ABC, CNN, and CSPAN television networks, and the NPR and PRI radio networks, with corresponding transcripts. This series provides training data for DARPA's "HUB4" project on continuous speech recognition in the broadcast field. This dataset was released by the University of Pennsylvania in 1997. Related paper: 1996 English Broadcast News Speech (HUB4) LDC97S44