To understand why long sentences are difficult to parse we need to understand how we generally parse sentences. In linguistics, often, parse tree are used to show how people structure sentences:
A parse tree is a tree structure in which sentences (S) are deconstructed in different phrases. There are
- noun phrases (NP), in which a noun (N) accompanied by a determinant (D) are stored.
- verb phrases (VP), in which a verb (V) and an object are stored.
The object can be many many things. As you can see in the below picture, another NP is added, to finish this simple sentence.
In this incredibly simple example, it is clear that we do not need to break down the sentence that much, nor is there any ambiguity. However, sentences can become more complex, by adding a prepositional phrase (PP):
Complexity and ambiguity
The longer sentences are, the more complex these trees become. The difficulty in creating these trees is finding the correct relations between phrases and, therefore, the structure of the tree. In some cases, even multiple sentences must be combined in one tree:
The second sentence is what comes below SBAR, the auxiliary node for a new S. Moreover, as you can see, there is a line leaving the right side of the picture. The sentence could thus become even longer and more complex.
When sentences are ambiguous, it is even possible to construct multiple trees that are correct, but have a different meaning. There are ways to notate these ambiguous trees, but I won't/cannot go into that. At linguistics.se, there is a question covering that.
Also, here is a post about creating parse trees/syntax trees in general, including a reference.