One central concern in linguistic research is the description of word order in grammatical sentences. However, for an adequate description of languages with a relatively free word order, previous formal means have proven to be less appropriate for these languages than for English:

Context-free grammars explode in size, which causes generalizations to be lost and grammars hard to write, maintain and process.

ID/LP grammars are inappropriate because they allow neither an ordering statement for a specific ID rule, nor may the phonology of a mother be anything other than some concatenation of the phonologies of its daughters. Yet, both kind of statements are empirically needed.

Feature logic grammars with appropriate expressional power do not suffer these shortcomings, but pose serious computational problems because the structural linguistic information is not explicitly separated from the remaining information. Explicit structural information would, however, guide the parsing process.

To overcome these problems, we propose the use of a linearization specification language (LSL) which is both empirically adequate and distinguishes structural information from the remainder to enable more efficient parsing.