Constituent marking has some limitations. First, the specified constituent borders are defined on the syntactic tree, not the dependency tree (dependency structures are an extra layer of annotation that is added to the syntactic structure). Using the tool therefore requires knowledge of the Alpino grammar and the syntactic trees that it generates.
Second, specification of the constituent type is necessary in most cases, especially for disambiguating prepositional phrase attachments. As shown in fig. 4, a noun phrase and a prepositional phrase can form a constituent on different levels. The two phrases can form either a noun phrase or a verbal projection with an empty verb (which is used in the grammar to account for verb second). The first structure corresponds to a dependency structure with a noun phrase internal prepositional modifier, the second corresponds to a dependency tree in which the prepositional phrase is a modifier on the sentence level. Marking the string het meisje in het park as a constituent without further specification does not disambiguate between the two readings: in both readings the string is a constituent. One has to specify that the string should be a noun phrase, not a verbal projection. This specification of the constituent type requires even more knowledge of the grammar. If one specifies a constituent type that cannot be formed at the denoted string position, the parser treats the specification as an illegal character, skips it and generates partial parses only.