Suppose we have sets A = {A1,A2.... An} and B = {B1,B2,...Bm}
and a relation R having elements like (Ai,Bj).
The aim is to make a partition set P of relation R such that each element (or cover of the partition) in the partition set is a cartesian product.
Example:
If R = {{A1,B1},{A2,B2},{A1,B2}}
then I want to make a partition set like
P = { { {A1,B1},{A1,B2} }, { {A2,B2} } }
Here both { {A1,B1},{A1,B2} } and {{A2,B2}} are cartesian products or we may say set of elements from A and B such that all elements Ai are matched with Bj in that set.
The union of these two sets make R and also { {A1,B1},{A1,B2} } and {{A2,B2}} are disjoint hence making P the partition set.
What algorithm/theorem should I follow to find a partition set like this? Note that R here will be a relation where always consisting of pairs as it's elements in which always first item will be from A and second from B. So {B1,A1} is not possible.
Other examples:
R = {{A1,B1},{A2,B2}} then P = {{A1,B1},{A2,B2}} because both {A1,B1} and {A2,B2} are independently cross products.
R = {{A1,B1},{A1,B2},{A2,B1},{A2,B2},{A1,B3}}
then P contains elements {{A1,B1},{A1,B2},{A2,B1},{A2,B2}} and {{A1,B3}}
You have a bipartite graph $G$ (with $A$ on one side and $B$ on the other, with edges determined by $R$). Then what you are asking for is known under the name biclique partition of $G$. You will find some literature about it online.
The problem of finding the smallest such partition (minimum cardinality biclique partition) is NP-hard, and I think it remains NP-hard in the class of bipartite graphs.