How are the tabular datasets preprocessed?

[BarclayII]

How are the tabular datasets preprocessed?

[alexis-cvetkov]

Hi, and thanks for your interest in our work!
Regarding your question, we do convert all tables to a graph, not only the single table with the entities of interest.
To that end, we manually select a column in each table, whose entries will be the head entities in the triples of the generated graph.

In practice this column often corresponds to the "main entities" of the table. For instance if I have two tables: one about students and one about courses, I will choose the columns "student" and "course".

Hope that's clear!