Semantic network

Revision as of 20:24, 18 September 2013

The main goal of the UNL is to represent, in a machine-tractable format, natural language meaning, i.e., the information conveyed by natural language documents. In the UNL framework, this information is represented by a semantic network, a network which represents semantic relations between concepts. This semantic network, or UNL graph, is made of three different types of discrete semantic entities: Universal Words, Universal Relations and Universal Attributes. Universal Words, or simply UW's, are nodes in the network; Universal Relations are arcs linking UW's; and Universal Attributes are used to delimit the use of UW's. This three-layered representation model is the cornerstone of the UNL, and a distinctive feature over other semantic networks, which normally propose only two levels: edges and vertices.

Universal Words (UWs)

UWs are expected to be associated to lexical realisation units in the UNL-NL Dictionary, which is a bidirectional bilingual dictionary mapping lexical items between UNL and NL. A single UW may correspond to several different natural language entries (synonymy), and one single open-class natural language entry may correspond to several UWs (homography). Entries from closed classes are not associated to UWs, but to relations or attributes. Numerals (such as "six", "sixth", "6"), formulae (H₂0) and untranslatable expressions (such as "http://www.unlweb.net") are represented as temporary UWs, i.e., they are not expected to be included in the UNL-NL dictionaries. The same happens to most proper names. Temporary UWs are automatically assigned the feature TEMP, and may be addressed by named entity recognition modules in UNL-based applications.

Universal Attributes

Universal Attributes normally convey information that may be directly associated to grammar categories, such as aspect, degree, gender, number, tense, mood, register, voice and social deixis. This association is made through D-rules, such as the following:

@pl = PLR;
@past = PAS;
@passive = PSV;
@male = MCL;
@past,@progressive = PAS,PGS;

Some attributes, however, cannot be directly assigned to any value, and are rather treated as features to be addressed by L-rules or S-rules:

(@ellipsis):=("",-@ellipsis); (L-rule)
(@square_bracket,%ref):= ("[")(%ref,-@square_bracket)("]"); (L-rule)
VC(@emphasis,%comp):=+IS(%comp)VC(%comp,TRACE); (S-rule)

Universal Relations

Universal Relations normally convey information that can be associated to S-rules:

agt(%source;%target):=VS(%source;%target);
tim(%source;%target):=VA(%source;PC([in];%target));