The Dictionary as a System of Concepts

In terms of componential analysis, the meaning of a word is a complex of semantic components (or features or markers) connected by logical constants. This assumption immediately allows us to define certain semantic properties and relations of lexical entries. Thus a word is semantically ambiguous if it has more than one complex of semantic features assigned to it. Two entries E and E are synonymous, if their meanings consist of the same components connected by the same logical constants. E is a hyponym of E (i. e. E is included in E) if the meaning of E contains the components occurring in the meaning of E, but not vice versa. The woman might be a hyponym of adult, since the former but not the latter contains e. g. the component female. E and E are antonyms, if their meanings are identical except that the meaning of E has a component С where that of E had C, and С and С belong to a particular subset of mutually exclusive components.

Sets oflexical entries whose meanings have certain features in common farm a semantic field. A famous example is that of kinship terms, whose elements share the feature configuration animate and human and relative. The verbs of motion are another example.

There are also subfields, e. g. that of male relatives including father, brother, son, uncle, etc., or that of lineal v. collateral kinship. On the other hand there are more inclusive fields, e. g. that of social roles, which includes, besides a subject of the kinship terms, elements like friend, colleague, teacher, etc. It follows therefore that semantic fields are relative, and not absolute. The notion of semantic fields was introduced by Trier in order to account for the observation that the meaning of lexical elements is specified only by their relatedness to and their difference from other relevant elements.

Besides having components in common, the elements of the vocabulary are connected to each other by other relations. One of these is the pertinence relation of which a particular instance is the relation between part and whole. The words arm, hand, finger, for example, denote parts of the human body, so that their meaning must contain a component that relates them appropriately to all entries whose meaning contains the feature human. Another instance of the pertinence relation is class membership. Thus an entry like member must be connected by a certain feature to words like set, class, club, party, in short to all entries shown by appropriate features to represent different types of sets or groups.

A different type of relation among the elements of the vocabulary is established by restrictions on their combinability. Thus verbs like talk, think, dream only allow subjects with the feature human; drink requires an object with the feature liquid; the adjective blond requires a subject specified by such appropriate features as human hair. Restrictions of this kind are called selection restrictions, as they indicate which lexical elements may be selected in order to form a semantically well-formed combination of two or more syntactically combined lexical elements. They specify, so to speak, possible semantic affinities among lexical entries. How these conditions are to be represented will be indicated briefly below.

In general, one might define a complex of semantic components connected by logical constants as a concept. The dictionary of a language is then a system of concepts in which phonological form and certain syntactic and morphological characteristics are assigned to each concept. This system of concepts is structured by several types of relations. It is supplemented, furthermore, by redundancy of implicational rules of the kind illustrated under (4), representing general properties of the whole system of concepts. This last assumption accounts for the fact that it is much easier to come to understand a new concept in a system whose general structure has already been acquired than an isolated element of a completely unknown system. At least a relevant part of these general rules is not bound to particular languages, but represents presumably universal structures of natural languages. They are not learned, but are rather apart of the human ability to acquire an arbitrary natural language. [...]