Section 5-2 SENTENCE INTERPRETATION
| A number of rules of interpretation appeared in the chapter three.
The linguist designed those particular rules to describe the interpretation of certain kinds of sentences as analyzed syntactically.
The rules gave interpretations for declarative, performative, imperative, and interrogative sentences as referring to propositions having characteristic defining semantic patterns.
|
|
Consider again the sentence diagramed in ¶5-1-6, figure 1: “This is John and this is Mary, but this is Bill.”
It consists syntactically in three independent clauses, which clauses for now we designate, CL1, CL2, and CL3.
These clauses happen to have the same type of interpretive structure.
|
| [+Declarative]CL1  r CAUSE(s,PERCEIVE1(h,TRUE(x))) |
| [+Declarative]CL2 r CAUSE(s,PERCEIVE1(h,TRUE(y))) |
| [+Declarative]CL3 r CAUSE(s,PERCEIVE1(h,TRUE(z))) |
| For the interpreter of the sentence, many of its features seem ultimately derived from a combination of features on its parts as dictated by its syntactic pattern.
Apparently the simplest features usually arise on terminal constituents which then combine and unify as the constituents group together syntactically.
The tree may then be seen as a semasiological map showing the order of combination and interpretation of features as inherited by the constituents of phrase structure.
An example may illustrate how S3 inherits the interpretation of CL3.
|
| [+Declarative]S3 [+Declarative]CL3 |
| [+Declarative]S3 r CAUSE(s,PERCEIVE1(h,TRUE(z))) |
|
There are a number of words in English that traditionally belong to the class of conjunction, which class we rename to grammatical connective (C).
The motivation is that logical conjunction is just one meaning possible for this class of the “conjunctions.” One connective is illustrated in the compound sentence under investigation — but, the meaning of which is adversative.
The other “conjunction” illustrated, and, is not.
Yet both these words include the logical meaning of the conjunctive operator AND( ).
It appears that this terminal constituent, the connective, has at least the following features: |
| [+Conjunctive]C | (for and, but) |  |
| [+Adversative]C | (for but) |  |
|
Linguists express the subcategorization of properties in terms of features by the use of segment structure rules.
We described in The Language of Science how features are designated and interpreted in a TG grammar.
Then in §4-5 the rules that generate segment structure were formally defined.
Originally linguists restricted this kind of rule to the terminal constituents of phrase structure.
As described and defined earlier it is much more useful to make them available to all constituents, the reason being that the rules may subcategorize the constituent syntactically, that is, that the syntactic environment,
i.e., information about the category, may be carried with the segment and inherited like the other semantic features.
When the constituent is terminal, its subcategorization is called “strict” subcategorization.
Inevitably the rules subclassify the category further.
Ideally the segment structure rules characterize all the semantics imposed by the language on the words and word patterns.
The specification of the possible features of a category is the ontology of the category as described to some extent for proper nouns in chapter 2.
|
| To make use of subcategorization for the part of speech that we called a conjunction, there is a rule like the first one in figure 3:
|
| Notice that the features in the second rule are discrete, i.e. individually distinct and mutually exclusive.
It is also possible to express features as Boolean being either present or absent.
Sometimes the features can each be assigned a number on a scale which names the whole set of Boolean features.
It is also possible for the semanticist using the language of logic to write a rule of interpretation for each separate connective to relate it to its appropriate semantic structure.
For example: |
| [+Conjunctive]C reference(C) = AND( ) |
|
There cannot be one rule to describe the interpretation of all sentences containing a connective.
The interpretation of the connective relates to the interpretation of the syntactic structures in which it is found.
In the case of the clause appearing in [P2] one needs to express the fact that the arguments of the relation AND( ),
which is a two-place predicate, are the content of the precMultiNeteding clause and the subsequent clause.
To do this we may enlist the services of a transformation.
In The Languages of Science the transformation called CONJ appeared as a rule of derivation in a TG grammar of the predicate calculus:
|
| CONJUNCTION (CONJ) |
|---|
| base1: | P1 |
| base2: | P2 |
| transform: | AND(P1,P2) |
| To keep things simple the linguist can build the rule for grammatical conjunction on the same pattern.
The following transformations adjust this form in two ways:
1) place the bases to the left as a single phrase structure, with the transformed phrase structure to the right, and
2) extend the relevant components using a broad line to its respective semantic interpretation in the predicate calculus.
|
| This means that the content of the superordinate clause is to be interpreted as the logical conjunction of the content of the clause preceding and the clause following the grammatical conjunction.
As a consequence of certain transformations, it is possible to reduce some of the syntactic structure that is generated by the PS-rules.
This is always done to conform the syntax of the output sentence to the logical structure of the sentence.
|
|
The linguist may describe the interpretation of but in a similar manner.
This rule requires an additional statement of the adversative relation that holds between the two propositions so conjoined.
An adversative relation means that the propositions asserted describe situations that are usually opposed.
One way to model this formally would be to refer to a fuzzy probability functor whose value lies between zero and one.
For example, “usual” is nowadays interpreted as around 80% probable (Hakel, 1968).
Presumably the conjunction but would be appropriate whenever the probability of the exclusive relation holding is greater than half.
|
| USUAL(x) probability(x) = 0.8 |
| PROBABLE(x) probability(x) > 0.5 |
| POSSIBLE(x) probability(x) > 0.0 |
![Adversative Conjunction of Clauses [Advers(CL)]](gif/ri/gramri14.gif)
|
The interpretation of the conjunction of the clauses in the sentence “This is John and this is Mary, but this is Bill” continues with reference to the PS in figure 3.
This is diagramed on the phrase structure (PS) map: |
| PS1: | [ ]CL11 [ CL111 [+Conjunctive]C11 CL112]CL11 |
by P2 and S1 |
| [ ]CL111 r y |
|
| [ ]CL112 r z |
|
| PS2: | [+Conjunctive]CL11 r AND(y,z) |
from PS1 by Conj(CL) |
| PS3: | [ ]CL1 [ [ CL11 [+Adversative]C1 CL12]CL1 |
by P2, S1, S2 |
| [ ]CL11 r AND(y,z) |
|
| [ ]CL12 r x |
|
| PS4: | [+Adversative]CL1 r AND(AND(y,z),x)  PROBABLE(EXCLUSIVE(AND(y,z),x)) |
| | from PS3 by Advers(CL) |
| Finally, the feature [+Declarative]S1 receives its standard interpretation presumably by the clause not being marked as of any other mode. |
| PS5: | [ ]S1 CL1 |
by P1 |
| PS6: |
| [+Declarative]S1r CAUSE(s,PERCEIVE1(h,TRUE(AND(AND(y,z),x) PROBABLE(EXCLUSIVE(AND(y,z),x))))) |
| from PS4 by Decl(S) |
|
In many cases it is possible to use the conventions of MultiNet to be explicit in describing the semantics of a sentence.
The designers’ goal in putting this system together was to represent knowledge.
Often they do not pay full attention to the syntactic constraints at work nor to assuring a direct correspondence between the syntax and semantics, especially with respect to grammatical conjunction,
where they keep things simple by omitting any explicit mention of it.
At the metalanguage level of description, conjunction is interpreted as equivalent to the inverted vee of the predicate calculus.
For MultiNet an assertion is simply the conjunction of all the atomic elements that go into making up the meaning.
Whenever conjunction must be mentioned explicitly as part of background knowledge, say, MultiNet uses one of its enumerator functors for the members of a set:
*ITMS for the pre-extensional elements and *ITEMS-I for the intensional elements.
If the members of the set are to be ordered in a series, the tuple functor is available: *TUPL.
Figure 5 illustrates our version of how MultiNet might describe the semantics of the sentence of figure 4.
|
| The use of MultiNet to describe the semantics of the adversative conjunction requires us to take advantage of the following observations.
Often the MultiNet definition of a relation fails to regard its possible atomic status.
In the case of adversative MultiNet generally uses opposition (OPPOS (x, y) and does not associate it with a measurement of probability.
Probability is designated explicitly as adverbial to the sentence and in English it is semantically a descriptor of a situation.
So the MultiNet expression would use 1) the *ITMS functor, 2) a specialization of the MODAL relation to the class of items relating to probability.
MultiNet expresses the meaning of EXCLUSIVE by using the functor *VEL2 (exclusive disjunction).
|
![Conjunction of Clauses [Conj(CL)]](gif/ri/gramri01.gif)
