training SimpLE model for link prediction on knowledge graph

Referring to this paper by Prof Kazemi, Prof Poole on SimpLE model for link prediction on knowledge graph.

In page 3, the paragraph on learning SimpLE Models, I understand that we have a batch of positive triples from the knowledge graph, where for each positive triple in the batch, we generate $$n$$ negative triple by corrupting it. So does that mean the batch size just increased by a factor of $$n$$ and there are only negative examples in the batch?

I think I understand the optimization function but I don't understand how we generated the labeled batch. Clarification would help!