The construction of dependency structures is driven by the lexicon. For each subcategorization type recognized in the lexical hierarchy a mapping between elements on the list-valued feature which specifies basic subcategorization properties (SUBCAT) and attributes of DT is defined. Two examples are given in figure 2. The leftmost feature structure exemplifies a finite, transitive verb. The value of DT of the nominative NP on subcat is identical to the value of the SU dependent. Similarly, the value of DT of the accusative NP on subcat is identical to the value of the OBJ1 dependent. The rightmost feature structure exemplifies a finite, transitive verb for which the object is assigned to the OBJ2 (secondary object) dependency relation. In some cases, the addition of dependency structures leads to more fine-grained distinctions. For instance, PP-arguments can be linked to PC ( prepositional complement) or LD ( locative or directional complement), where the distinction between these two is primarily semantic in nature. Therefore, verbs taking a prepositional complement are assigned a subcategorization frame that differs from the frame assigned to verbs taking such a LD complement.
In HEAD-COMPLEMENT structures, the DT attribute can simply be shared between head daughter and mother. In HEAD-MODIFIER structures, the dependency structure of the modifier is added to the list-valued MOD dependent of the head.