The problem is, that a lot of definitions for this term, not including musings by people high on intellectual (and sometimes chemical) hallucinogens are available. This is an obstacle, because any discussion on the topic begins with a bias from where the individual obtains his version of its definition.
My opinion (not just a vagrant, unsubstantiated one) is, that all such important concepts be defined using laws of parsimony coupled to self-compatible and consistent ideas which are subject to change with availability of new knowledge and is explicitly accepted as such.
That sounds complex. Fortunately it is not. Contrary to popular point of view the nature of reality is not a matter of choice, especially when definitions are made to suit such needs. 'You tend to fall when you jump from high altitudes', being an example of its unchanging nature. Its the interpretation of its nature that is variable. All we are sure of, are facts such as one described earlier, everything else is conjunctures based on interpolation and assumptions. This means that today (and perhaps in distant tomorrow as well) all we have is a limited information that is considered fact. The collection of these solely is NOT reality.
Definition: Reality is a collection of facts associated with a consistent model.
A model is an explanation that gives the best possible approximation (I would say explanation, but what the hell!) towards interconnection of different facts, that are seemingly experimentally (empirically) correlated. Thus, when we garner new facts we have to recheck the consistency of current model. If contradictions are found, we improve the current model with a consistent one. It may now happen, that several models are available to justify a given collection of facts. In that case, we test the justifiability of all the models by closely inspecting what they predict and checking which predictions turn out to be correct.This process is progressive. There always will be more than one model to explain a given collection of facts and thus always an investigation into the correctness of every model; thus producing in all likelihood new facts.
Now that you have noticed, empirical evidence is given a special preference here in the essay to be a method to explore reality. There is a good reason for that. Empirical facts don't change with time, given a context. That is actually the definition of an empirical fact. They serve as standards to compare against, which makes them terribly useful in the entire process. There is no doubt that purely intellectual (used as opposed to empirical) methods can and do generate knowledge. But that's exactly what models are. So we have a meaningful co-existence of either methodology.
Finally, there is one thing that remains unexplained here: the expression of the model.
Colloquial languages are are not made; they are grown. Which is to say that they evolve independent of a strict, rigorous logic. By contrast, Mathematics is a language that has been made to be axiomatic and rigorous for the purpose of explaining the inter-relation between collection fo facts. Thus it suits best the purpose of modeling the nature of reality.
Undoubtedly this essay is a simple primer. But sometimes it is necessary to write these things down for the sake of mental clarity.