A time machine for information: Looking back to look forward

Historical data (also called long data) holds the key to understanding when facts are true. It is through long data that one can understand the trends that have developed in the past, form the audit trails needed for justification, and make predictions about the future. For searching, there is also increasing interest to develop search capabilities over long data. In this article, we first motivate the need to develop a time machine for information that will help people look back so as to look forward. We will overview key ideas on three components (extraction, linking, and cleaning) that we believe are central to the development of any time machine for information. Finally, we conclude with our thoughts on what we believe are some interesting open research problems. This article is based on the material presented in a tutorial at VLDB 2015.