We present FactWatcher, a system that helps journalists identify data-backed, attention-seizing facts which serve as leads to news stories. FactWatcher discovers three types of facts, including situational facts, one-of-the-few facts, and prominent streaks, through a unified suite of data model, algorithm framework, and fact ranking measure.
Acknowledgement: This work is partially supported by NSF grants IIS-09-16027, IIS-10-18865, CCF-11-17369, IIS-13-20357, and IIS-14-08928. Additional support comes from HP Labs Innovation Research Awards, a Google Faculty Research Award, and National Natural Science Foundation of China Grant 61370019. Any opinions, findings, and conclusions in this work do not necessarily reflect the views of the funding agencies.
Disclaimer: The NBA and Weather datasets are collected from www.databasebasketball.com and data.gov.uk/, respectively. The NBA dataset has the box scores of NBA games from season 1991-92 to 2004-05. There are known errors in the original dataset which cannot be fixed without proper information sources. Moreover, the dataset does not include records of all seasons. The weather dataset has the weather measures of more than 5000 locations across the United Kingdom. There exist dirty data problems. We did not attempt to fix those. Hence, the facts generated by our system do not necessarily stand in the real world.