Universal Networking Language

From Wikipedia, the free encyclopedia

UNL can also mean University of Nebraska-Lincoln.

In machine translation, Universal Networking Language (UNL) is an artificial pivot language, that relies on the semi-automatic translation from the initial text in a natural language into its pivot equivalent. Documents expressed in the pivot language can then be automatically translated into another natural language.

The pivot representation consists of concepts and relations between them. Concepts are not formally defined but represented by their English equivalent. Allowed concepts are listed in the Universal Word Lexicon (UW concepts). The relations between the concepts are drawn from a set of predefined relations.

The Universal Networking Language Project is an attempt to build a knowledge-based translation system. It has developed a format specification of an intermediary language (UNL), a limited lexicon (UW KB) and several demos each of which converts UNL to a natural language. As of 2004, the project is not finished yet.

BabelCode, UNL and KCE (by Carnegie Mellon University) are all more or less based on knowledge representation theory but apply it in different levels.

1 Critical Remarks
2 Who created UNL?
3 See also
4 External links

[edit] Critical Remarks

UNL is not the first attempt to use a pivot for machine translation. These systems, called interlingual machine translation system worked, if at all for a limited domain with a limited number of concepts and a limited number of relation between the concepts. Trying to list all concepts of all cultures seem in this light an impossible task.

The technique to use an English word to represent a concept is also questionable. First of all, many concepts of other cultures cannot be expressed in one English word. Other languages make distinctions English does not make lexically. Second, most English words have more than one meaning, e.g. construction, standing for the process and the result of a process. Just giving an English word to define a concept without making the meaning distinction is insufficient. Third, many different English words have the same meaning, e.g. to construct, constructing and construction (in the process reading). Listing them in a dictionary of concepts creates concepts which cannot be distinguished from other contexts. Forth, the question whether two English words express the same concept is context depended. In a chemical text, water and H2O might refer to the same concept. In other texts they might not.

Thus, the claim of UNL to be useful for every language can be seriously questioned. Even if some simple test have been run with two handful of languages, the diversity and richness of human languages has been barely explored.