Computational grammars for English, which are based, more or less directly, on linguistic theories are the Alvey [19,45] and CLE  grammars (both based on GPSG), XTAG (based on TAG ), the Parallel Grammar Project at Xerox (based on LFG, for English, French and German ), and ERGO  and ALE  grammars (both based on HPSG). Typically, such grammars account for phenomena such as subcategorisation, agreement, auxiliary inversion, copula and small-clause constructions, passives, long-distance dependencies, relative clauses, extra-position, PP-attachment, anaphoric binding, adverb placement, etc. The implemented grammars are intended as faithful, yet computationally feasible, implementations of a given linguistic theory. Linguistic processing of such systems requires, at least, the ability to handle complex data-structures (usually (typed) feature structures), under-specification and unification, highly under-specified syntactic schemata (in which the order and number of the daughters may be under-specified), (recursive) lexical rules, and, sometimes, the ability to handle default mechanisms of various kinds.
Below, we propose to develop a linguistically motivated computational grammar for Dutch, comparable in coverage to systems available for English. At the moment, there is no such grammar available for Dutch, although a non-trivial grammar fragment has been developed within the NWO Priority Programme Language and Speech Technology (TST).
Note that developing such a grammar will involve implementing accounts for a number of phenomena which do not occur, or occur to a lesser extent, in English, such as cross-serial dependencies, the word order differences between (verb-final) subordinate clauses and (verb-initial or verb-second) main clauses, and the relative free order of adjuncts in the ``Mittelfeld''.
We propose to take the TST fragment as a starting point for a more general grammar. Before we explain what is involved in the development of such a grammar, we will first motivate this choice as follows.