The lexicon

The lexicon is a list of clauses lex(Word,Sign), associating a word (or sequence of words) with a specific sign.

Constraint-based grammars in general, and lexicalist constraint-based grammars in particular, tend to store lots of grammatical information in the lexicon. This is also true for the OVIS2 grammar. A lexical entry for a transitive verb, for instance, not only contains information about the morphological form of this verb, but also contains the features SC and SUBJ for which quite detailed constraints may be defined. Furthermore, for all lexical signs it is the case that their semantics is represented by means of a feature-structure. This structure can also be quite complex. To avoid massive reduplication of identical information in the lexicon, the use of inheritance is therefore essential.

Figure 3: Fragment of the lexical hierarchy

In figure 3, we illustrate the use of inheritance in the lexicon. All lexical entries for verbs have a number of properties in common, such as the fact that they are of type v, and take a normal (non-locative and non-temporal) NP as subject. This is expressed by the template v(V). Intransitive verbs ( iv(V)) can now be characterised syntactically as verbs which do not subcategorise for any (non-subject) complements. Transitive verbs ( tv(V)) subcategorise for an NP with accusative case. The templates intransitive(Pred,Sign) and transitive(Pred,Sign), finally, combine the syntactic and semantic properties of intransitive and transitive verbs. The variable Pred is used in the semantics to fix the value of the predicate defined by a particular verb. A limited form of non-monotonic inheritance is supported (see Carpenter [15] and Bouma [12] for more general approaches). For instance, `weather' verbs require the dummy pronoun het ( it) as subject, but behave otherwise as intransitive verbs. This can be expressed by letting weather_v inherit from iv, with the exception of the value of the NFORM attribute of (head of the list containing) the subject, which is assigned an exceptional value. The attribute-value matrices for the templates iv(V) and tv(V) are:

$$\small\begin{minipage}[t]{.9\textwidth}\begin{avm} \mbox{\tt ... ...& $\langle\rangle$ \end{displaymath}~). \end{avm}\end{minipage}$$ (10)

The lexicon itself (i.e. the predicate lex/2) is defined in terms of the predicates entry, inflection and lexical_rules:

$$\small\begin{minipage}[t]{.9\textwidth}\begin{verbatim}lex(... ...Sign1), lexical_rules(Sign1,Sign).\end{verbatim}\end{minipage}$$ (11)

The definition of entry(Root, Sign) defines for each root form what its associated sign is. For instance, for verbs we must typically distinguish a first person singular form, a second and third person singular form, and a plural form (which is also the form of the infinitive). The predicate inflection defines how inflected forms are derived. For example, there is an inflection rule which adds a t to the base form of a verb, and specifies that its agreement features are third person singular, and its VFORM value is fin. Lexical rules can be used to transform the sign associated with a lexical entry. For instance, the account of nonlocal dependencies sketched below makes use of a lexical rule which removes a sign from SC and places it on SLASH. A more detailed account of this lexical rule is given in the section on nonlocal dependencies. As an example, assume the stem arriveer (to arrive) is defined as an intransitive:

$$\small\begin{minipage}[t]{.9\textwidth}\begin{verbatim}entr... ...n):- intransitive(arriveren,Sign).\end{verbatim}\end{minipage}$$ (12)

Such a definition will give rise to a number of lexical entries. One of these will be the third person singular finite form:

$$\small\begin{minipage}[t]{.9\textwidth}\begin{avm} lex(arrive... ...& $\langle\rangle$ \end{displaymath}~). \end{avm}\end{minipage}$$ (13)