The authors of this paper in their previous work used partial ordering to model the relationship between reading materials. These are the rules that were used:

1. A practice problem (a reading text) can be characterized as a multiset of its required concepts.
2. Problem s1 is harder than problem s2 (indicated as s1 > s2) if and only if s1 covers all required concepts of s2. This also implies that students who understand s1 will also be able to understand s2.
3. Problem s1 is directly harder than problem s2 if s1 > s2, and there is no other problem s3 such that s1 > s3 > s2.
4. A partial ordering graph is a Direct Acyclic Graph (DAG): each node represents a practice problem and each edge represents a “directly harder than” relation between two problems.

In order for this algorithm to work the hierarchical structure of domain knowledge should dense, otherwise there will be too few edges. That's the case for the vocabulary knowledge. In order to increase density, new rule is applied:

Problem s1 is $\alpha$-fuzzily harder than problem s2 if s1 covers at least a proportion $\alpha$ of required concepts of s2. Using this fuzzy partial ordering, we can also define the fuzzy partial ordering graph.

The $\alpha$ parameter should be chosen wisely because if it is too small there would be too many edges and the likelihood that student would understand the problem would be too low if it is too high there would be too few. The authors set $\alpha$ to 0.8