19.3.1 Dependency Grammar

A dependency grammar G is a 7-tuple

$\langle \SETWORDS, \SETCATS, \SETAGRS, \SETCOMPS, \SETMODS, \SETLEX, \SETRULES\rangle$

$\SETWORDS$: a finite set of strings notating the fully inflected forms of words
$\SETCATS$: a finite set of categories such as det for determiner or vfin for finite verb
$\SETAGRS$: a finite set of agreement tuples such as <masc sing 3 nom>
$\SETCOMPS$: a finite set of complement role types such as subject or dative (for dative NP complement)
$\SETMODS$: a finite set of modifier role types, such as adj for adjectives, disjoint from $\SETCOMPS$ . We write $\SETROLES=\SETCOMPS\uplus\SETMODS$ for the set of all role types. They will serve to label the edges of a dependency tree.
$\SETLEX$: a finite set of lexical entries (see later)
$\SETRULES$: a family of binary predicates, indexed by role labels, expressing local grammatical principles: for each $\rho\in\SETROLES$ there is a $\Gamma_{\rho}\in\SETRULES$ such that $\Gamma_{\rho}(w_1,w_2)$ characterizes the admissibility of an edge labeled $\rho$ from mother to daughter .

In practice, in addition to category and agreement, we will need other features to encode miscellaneous details (e. g. `zu' particle, detachable verb prefixes, choice of auxiliary verb). For the time being, we will simply ignore them and postpone their introduction until the presentation of our implementation.