| The linguist has formulated phrase structure rule [P1] and [P2] to generate more than one clause as extensions to the sentence. There are additional phrase structure rules that come later in the sequence, viz., [P4], [P5], [P7], [P12], [P19], and [P20], for introducing a dependent clause. Phrase structure rules can characterize the structure of very complex sentences, but are incapable of describing the co-reference of their various parts, which often manifests itself in parataxis as anaphora. |
Transformations to describe anaphora.The linguist may, however, appeal to grammatical transformation to describe the anaphoric aspects of the structure of a complex sentence. Transformations allow the analyst to combine multiple sentences each of which has been described with phrase structure rules. The transformations must have important restrictions on their applicability. To describe anaphora they must state a condition — that there be co-reference of a certain kind. Appropriate parts of the two sentence structures from which the resultant sentence structure derives must be interpreted as referring to the same thing. |
The imbed and the matrix.The kind of transformation needed to describe the relationship between a dependent clause and the independent clause of which it is a part starts out with at least two bases — one for each clausal structure. The linguist calls the first base the imbed. This base describes the structure of the dependent clause. Another base, the matrix, describes the structure of the main clause. The reason for saying that appropriate transformations apply to each dependent clause is to assure that the complex sentence conforms to some standard of grammaticality and can be interpreted. |
| Figure 11 illustrates some of the elements of analysis that go into building the two basic kinds of paraphrase and the rules of interpretation involved. In the first case each assertion is represented by an independent clause and compounded together as co-equals syntactically. When there is a connective present, it will suggest the kind of relationship that stands between the two assertions. In addition there may be some semantic dependence imposed by the order that should be accounted for in the rule of interpretation. This kind of behaviour proliferates the number of such rules and begs for their subclassification based on the meaning imposed. In the second case the clausal parts of the sentence are syntactically dependent. There are many rules of interpretation each classified by how the dependent clause relates syntactically to the independent clause. The imbed of one matrix may be the matrix of another imbed. Sometimes such clauses as stand in both relationships are elided so that evidence of their existence comes from the nature of the connective and the complex meanings conveyed by it. |
The strategy of MultiNet.MultiNet does not describe anaphora, but chooses simply to represent its interpretation at the pre-extensional level. The syntactic constructions of English (and German) are part of the semantic description in that the distinctions of figure 11 fall into the ontology of situations. In the case of simple conjunction, the system describes its parts as elements of the representation of native knowledge. Hypotaxis is folded into the 52 relations and functors whose signatures (domain and range) may include situations. These fall into four main kinds: |
| 1) seven relations for describing spatio-temporal embedding and a family of functors for representing the prepositions of locality |
| 2) one relation for describing modality and a functor for negation |
| 3) nine relations for describing hypotaxis between situations, such as reason and condition, and a family of functors for exclusive disjunction and one each for ordered and unordered conjunction, each member distinguished by the number of elements disjoined |
| 4) the many relations for describing some of the important ways in which objects may be involved in a situation. This last class includes 16 relations and one functor for the accompanying circumstances described in adverbial clauses as well as 15 for the participants in states and events. |
