Universality of semantic frames versus specificity of conceptual frames

FrameNet is a semantic network that links semantic frames, each evoked by a set
of lexical units and consisting of frame elements (with semantic types, definitions
and relations) that outline the semantic structure of the frame, as well as
frame-to-frame relations and annotations that illustrate the syntactic realisation of the
frame elements.

The Bulgarian FrameNet is based on the FrameNet and at the same time offers the
possibility to encode language-specific semantic structures, either by replicating or
reconstructing existing semantic frames or by introducing new frames. An abstract
representation, called superframe, is developed to replicate language-independent
information (at least for English and Bulgarian) from semantic frames. In the
Bulgarian FrameNet, the conceptual frames inherit either all or part of the language-
independent information from the semantic frames via the superframes and may
contain additional language-specific data to represent scenarios evoked by the
Bulgarian lexical units. Each conceptual frame is extended by a set of nouns that
represent the lexical realisations of the frame elements corresponding to the target
lexical units.

The study presents the FrameNet, the creation of FrameNets for other languages,
the motivation for introducing the conceptual frames and the superframes, which
combine the semantic and conceptual frames in a multilingual network, and the
structure of the Bulgarian FrameNet, which includes Lexical, Grammatical, Frame
and Syntactic sections (valence patterns).

The overall aim is to present our approach to the identification and transfer of
language-universal knowledge from the FrameNet semantic frames, universal in
the sense that it applies to both English and Bulgarian, and the definition and
integration of language-specific components of the conceptual frames for Bulgarian
(as compared to English).