Semantic rules

From UNL Wiki
Jump to: navigation, search

Semantic rules are rules for mapping UNL structures into natural language and vice-versa. They can be of two different types: relation rules, i.e., rules that map UNL relations into syntactic structures; and attribute rules, which are rules that map UNL attributes into morpho-syntactic structures.

Contents

Relation Rules

SEMANTIC RELATION := SYNTACTIC RELATION;

Relation rules are used for translating UNL relations into natural language syntactic structures. They form a special case of the S-Rule type, because:

  • differently from other s-rules, the heads of the syntactic relations may not be simplified; and
  • differently from other s-rules, both arguments of the syntactic relation must be coindexed to the arguments of the semantic relation through the placeholders %01 and %02, which stand for the left and the right argument [of the semantic relation], respectively.

Templates

CASE TEMPLATE* EXAMPLE GLOSS
ADJUNCT (DIRECT) XA(%01;%02); mod(NOU;ADJ):=NA(%01;%02); if "mod(NOU;ADJ)" then the ADJ (%02) is a direct adjunct to the NOU (%01)
ADJUNCT (INDIRECT) XA(%01;YC(Y;%02)); tim(VER;NOU):=VA(%01;PC([in];%02)); if "tim(VER;NOU)" then the NOU (%02) is the complement of the preposition "in" which is the head of the prepositional phrase adjunct to the VER (%01)
COMPLEMENT (DIRECT) XC(%01;%02) obj(VER;NOU):=VC(%01;%02); if "mod(VER;NOU)" then the NOU (%02) is the direct complement of the VER (%01)
COMPLEMENT (INDIRECT) XC(%01;YC(Y;%02)); gol(VER;NOU):=VC(%01;PC([to];%02)); if "gol(VER;NOU)" then the NOU (%02) is the complement of the preposition "to" which is the head of the prepositional phrase complement to the VER (%01)
SPECIFIER XS(%01;%02) agt(VER;NOU):=VS(%01;%02); if "agt(VER;NOU)" then the NOU (%02) is the specifier of the VER (%01)

*X and Y must be replaced by lexical categories (N,A,J,V,P,C,D,I), depending on the nature of the heads.

Attribute Rules

Attribute rules are used for translating UNL attributes into natural language morpho-syntactic structures. They can be of three different types:

Feature assignment

ATTRIBUTE := TAG;

To be used when the attribute is equivalent to an existing natural language feature

  • @pl := PLR;

the attribute .@pl, which stands for plural, is directly equivalent to the tag "PLR"

  • @past := ET0;

the attribute .@past, which stands for past, is directly equivalent to the tag "ET0"

  • @past.@progress := ET0&RT1;

the attribute .@past.@progress, which stands for the past continous, is directly equivalent to the tagset "ET0&RT1"

m-rule

ATTRIBUTE := ADD < DELETE; (PREFIXATION)
ATTRIBUTE := DELETE > ADD; (SUFFIXATION)
ATTRIBUTE := DELETE : ADD; (REPLACEMENT)

To be used when the attribute cannot be associated to an existing tag and implies a morphological or orthographical change

  • @interrogative := 0 > "?";

the attribute .@interrogative implies the generation of a question mark "?" at the end of the sentence

  • @square_bracket := "[" < 0, 0 > "]";

the attribute .@square_bracket implies the generation of an opening bracket in the beginning of the sentence, and of a closing bracket in the end of the sentence

s-rule

ATTRIBUTE := SYNTACTIC RELATION;

To be used when the attribute cannot be associated to an existing tag and implies a change in the syntactic structure of the sentence (i.e., the insertion of a word)

  • @def := NS(DP([the]));

the attribute .@def implies the generation of a definite article in the position of the specifier of the noun

Software