To answer that question I will cite the “Scope and goals” section of the Wikipedia article
and add in line comments where my approach diverges. Normal font for common parts [italic for parts specific to UNL] [bold for parts specific to my approach].
UNL is designed to establish a simple foundation for representing [the most central aspects of] [, without preset constraints,] information and meaning in a machine- and human-language-independent form. As a language-independent formalism, UNL aims to code, store, disseminate and retrieve information independently of the original language in which it was expressed. In this sense, UNL seeks to provide tools for overcoming the language barrier in a systematic way.
At first glance, UNL seems to be a kind of interlingua, into which [source texts are] [live input is] converted before being translated into target languages. It can, in fact, be used for this purpose, and very efficiently, too. However, its real strength is knowledge representation and its primary objective is to provide an infrastructure [for handling knowledge that already exists or can exist in any given language] [that enables much easier aquisition and creation of knowledge, and constructive discussions, that crosses the present day cultural and academic boundaries].
[Nevertheless, it is important to note that at present it would be foolish to claim to represent the “full” meaning of any word, sentence, or text for any language. Subtleties of intention and interpretation make the “full meaning,” however we might conceive it, too variable and subjective for any systematic treatment. Thus UNL avoids the pitfalls of trying to represent the “full meaning” of sentences or texts, targeting instead the “core” or “consensual” meaning most often attributed to them].
[I treat the context free “full meaning” of words as a mostly a meaningless abstraction that is not worth pursuing as an aim. The only pertinant meaning to me is the one in the authors mind. The aim is to make that meaning available to the readers with the least possible deformation. The unit of meaning I propose is a word sense that includes its context (or domain of validity).
When the author inputs a text into the system he either finds one in the system or creates it on the spot. That creation will be part of the text, so there is no need of a centrally managed ontology. As in natural language such “neologisms” propagate with their usage. They can be included in independently managed ontologies, that can specialise either in general or specialised domains, and made available to authors].