Discourse Representation Structure

Discourse Representation Structures (DRSs) can be displayed in different formats. In the PMB we use three different formats: the sequence notation (for annotation and evaluation purposes); the graph notation (for visualisation purposes); and the box notation (as introduced in DRT).

Sequence Notation

The sequence notation is a way to represent DRS without using variables or brackets. The variables are replaced by indices that refer backwards or forwards in a sequence of concepts. The concepts are aligned to the linguistic input with a percentage symbol (anything on the right of a percentage sign on a line does not belong to the meaning representation). The sequence notation has the following ingredients:

  • Concepts (cat.n.01, see.v.03, ...): always written in lowercase, denoting a WordNet triplet consisting of a lemma, part of speech ("n", "v", "a", or "r") and a sense number consisting of two digits;
  • Constants ("Mary", speaker, 20, π, ...): denoting proper names (of people, animals, organisations, locations, artifacts), numerical values (integers and reals), times/dates, and deictic references (speaker, hearer, now);
  • Roles (Agent, Theme, Patient, ...): connecting events with entities, always written with an uppercase letter followed by lowercase letters;
  • Operators (=, ≠, ≈, <, ≤, ≺, ...): written with three uppercase letters;
  • Indices (..., -2, -1, +1, +2, ...): selecting the external argument of a role;
  • Contexts: a sequence of zero or more concepts;
  • Separators (NEGATION, CONJUNCTION, EXPLANATION, CONTINUATION, ...): introducing new contexts via a discourse relation - written in all uppercase letters;
  • Connectors (..., <2, <1, >1, ...): selecting the context to which the new context attaches to;
A role followed by an index is a hook. A role followed by a constant is an anchor. A simple sequence is a sequence of one ore more concepts, where a concept can be followed by one or more hooks or anchors. A simple sequence represents a single context. A context is similar to a box in DRT. A complex sequence is formed by combining two (simple or complex) sequences using a separator and connector.

Graph Notation

A meaning in sequence notation can be visualised as a directed acyclic graph, where the vertices denote concepts, contexts or constants, and the edges are decorated by roles or comparison operators. Concept nodes are drawn as ovals and context nodes as boxes. The NEGATION separator is visualised by ¬ and the CONJUNCTION separator is eliminated for readability by merging the two contexts that it connects.

Box Notation

The box notation as introduced in DRT, with variables restored. Each box (DRS) corresponds to a context in the sequence notation. Each concept in the sequence notation introduces a new discourse referent.