| As with a TG grammar just constructed, it is also possible to build a calculus for the description the semantic features used by knowledge engineers and linguists to represent the meaning of natural languages. Here we proceed in a way analogous to TG grammar. These principles were used in our introduction to the phonology of English. There we demonstrated the way such a grammar can be used to describe the stucture of some of the various sounds of language. When this is done linguists are able to relate the structure of each segment as a whole to a constituent of syntax. The constituents thereby become complex symbols. From the point of view of syntax each constituent is a part of speech occurring in various patterns. From the point of view of semantics it is a bundle of features that in patterns may be unified with other bundles in the process of language interpretation necessary for communication. |
| So far as the most basic elements of meaning are concerned, linguists have done their best to identify some minimal required to describe all cultures. However, as with syntax and phonology, the analyst will find that it is the native elements and structures of English that take center stage in constraining the form that the rules of the semantic component take. This section extends their efforts by introducing the tools of a formal approach designed to describe the meanings of syntactic units of various kinds and sizes for both logical and natural languages. |
Describing segment structure.The linguist as knowledge engineer defines a constituent as describable in terms of a segment structure ( ), an ordered sequence of features ( ).
Box one contains two formation rules in (1) and (2) to describe the calculus that in turn will enable a description of the linguist’s grammar.
|
| In accord with the second rule, suppose that (3) is the segment structure of some constituent of English syntax. A convention we have adopted is to represent features as labels in yellow boxes and then draw boxes around the features that belong together as a feature chain. Because particular bundles generated by the the rules may be relatively large, we have also developed ways to abbreviate them. Thus, when a feature is placed in a tan box, the intent is that it be an abbreviation or a particular chain as defined by rule. The form in (5) is the way some computer scientists diagram classes. This is another of the tools established in the Unified Modelling Language (UML). |
Rules to describe segment structure.As with the calculus for syntax, the way to get from the formation rules to the segment structure is by means of a set of segment structure rules ( ).
These most generally have the form in (1) in box 8.
The structures characterized in (2) and (3) are generated by (2) of the calculus in box 7.
The SS-rule in (1) has the effect of extending the feature chain that describes the segment as distinctive from all other segments.
|
| This rule is part of the TG grammatical apparatus. This is a replacement rule for interpreting or deriving one structure from another by rule. Given a structure as in (2) and the rule in (1), this tool derives the structure in (3). Linguists often avoid the use of an environment structure. The reason is that they may usually replace such a condition with a suitable transformation. When the feature added to the chain is not distinctive, i.e., it cross-classifies with other segments, the environment is important for distinguishing the segment. The “context-free rules” will then need to have the context retained in a special feature called a rule of “strict subcategorization.” It is quite common for the terminal constituent of phrase structure to be thus distinguished by its context. |
Abbreviating segment structure rules.It is most useful to abbreviate rules multiple rules with identical left sides and similar right sides. This illustrated, with its own rule of interpretation as (2) in box 9. One further step in abbreviating the set of SS-rules is illustrated in the ontology chart, such as introduced in §§ 2-3, 2-4. On this kind of chart the cascading nature of classification by rule is made clear using curly brackets. Normally there are only some limited number of rules at each level of classification. In these cases the features may be indexed, i.e. ordered by number. Some features are binary, others are scalar, or n-ary. |
As in the rules of phonology certain bundles of features are conveniently assigned additional new features that can be predicted by virtue of the bundle.
In semantics these are default attributes.
For example, the feature of  is assumed to be present when  is, since plants are generally constrained to one spot.
This is done by means of redundancy rules as characterized in box 10.
In MultiNet such rules are called B-axioms.
There are a couple of other kinds of abbreviations with their interpretations given in the box.
One feature may stand for many, for example, the feature  , which we introduced in chapter 2.
Co-occurring segments may be combined or a single segment may be split apart.
These abbreviations are most useful in morphology, but also find a place in syntax.
|
Negative feature convention.Sometimes it is important to characterize a segment by what feature it fails to contain. This is easy enough if the feature is binary. In case the feature is scalar, a segment containing any of its other values should be specified. Such a feature is illustrated in box 11. The ad hoc feature needs a name, which with this convention is the name of the feature that it opposes together with the “non-” prefix attached. As a further reminder that this feature stands for any one of the other values, the yellow background color is stippled. The actual value of the feature fitting this criterion is a negated exclusive disjunction of the features possible. |
Exclusive vs. inclusive disjunction.Another useful characterization of a segment is by multiple features that fail to be present. In this case care must be taken since the simple exclusive set of values not present for one includes the others. Box 12 shows the kind of logical symbology that reminds us that the complement of multiple features must be the intersection of the complements of each one. This is called negated inclusive disjunction. |
