The dataset consists of 90 million tables out of the 233 million Web tables in the corpus. In relational tables, a set of similar entities is described with one or more attributes.