The primary reference describing the SIL semantic representation of discourse is [Beekman et al 81]. They break meaning representation into three levels: concepts, propositions, and propositional clusters.
The constituents of a concept are meaning components. For instance, they give the following analysis of the verb ``to run'':
Of these six components, the first is the central, nuclear one. Each of the others is related to the central component by the general relation they call ``delimitation'', that is, they narrow down the very general central component until the combination yields the specific concept desired. Concepts can be assigned to one of four classes: things, events, attributes and relations.
Propositions are divided into state propositions and event propositions. Event propositions use a fairly straightforward case grammar analysis. State propositions, however, are given an interesting analysis:
type example statant state naming My dog is called ``Fido'' my dog Fido identification That is a wrench that wrench description He is sick he sick substance That is made of wood that wood depiction The story is about Bill the story Bill ownership That car is mine that car mine partitive The tree has a trunk tree trunk location The car is in the garage car garage containership This bag contains rice bag rice derivation Milk comes from cows milk cows kinship role She is my sister she my sister social role She is a doctor she doctor function This room is for storage this room storage classification A dog is an animal a dog animal ambience It is hot (ambience) hot time It is noon (time) noon existence There is evidence evidence (there is)
These classifications of states was found to be very useful.
The analysis of propositional clusters in this project was taken from the SIL system presented in [Beekman et al 81]. This system was applied to the analysis of much of the New Testament in publications called ``Semantic Structure Analyses''. These publications will be used as the basis for encoding the semantic structures of the source text.
In general, propositions are seen to be connected into propositional clusters using a system of relations, or rhetorical predicates. [Beekman et al 81] defines the basic system as follows:
Addition relations
chronological
sequential
simultaneous
non-chronological
conjoining
alternation
Support-head relations
chronological
non-stimulus-response
orientation
circumstance
time
preliminary incident
closing
progression
step1-step2...stepn goal
stimulus-response
speech
evaluation
question execution
remark answer
proposal reply
counter
non-speech
occasion-outcome
problem-resolution
non-chronological
orientation
orienter-content
opening-HEAD
HEAD-closing
introduction-HEAD
restatement
HEAD-equivalent
HEAD-amplification
generic-specific
clarification
HEAD-illustration
HEAD-manner
contrast-HEAD
logical
reason-result
result-means
means-purpose
condition-consequence
concession-contraexpectation
HEAD-grounds
A text is analyzed by breaking it into a hierarchical tree, where each node is one of the relations above and the leaves are the actual propositions found in the text. Each of the support- head relations has nuclear part (the head) and a supporting part. This head-support relation was very useful in the focus tracking methodology introduced in this project.
Steve Beale