>I see that the XML document is often represented as tree structure.
I guess it should not be "often". I would say you can say it more catergoric without "often", meaning "always".
>I see that the XML document is often represented as tree structure. But in some papers it is a unordered tree.
I would say, as a matter of understanding, the standard is not restricted it to any type. The ramification, a bit naive here, is that for a pure ordered xml document tree, you are restricted to only sequence content particles in its dtd definition, like this.
[tt]<!ELEMENT w (x,y,z)>[/tt]
However, for purely unordered tree type, the validating dtd would be restricted to only choice content particles.
[tt]<!ELEMENT s (t|u|v)>[/tt]
Besides you could equally ask why there is not n-ary restriction too. But, all these restrictions would be too severe for real-world app and non-abstract-thinking average "professionals".
It is a good topic for research though I would guess. As you might know ordered tree can be mapped to a binary tree. And also the complexity of an enumeration of unordered tree is not more than of an ordered tree.
Hence, what I would say as to "why" is that the standard is not drafted with 100% freedom. It must be a trade off among easiness of use and understood and implementation, and rigor.