ID/LP rules

Generalized Phrase Structure Grammar (GPSG) uses ID-rules (immediate dominance rules) and LP-rules (linear precedence rules) instead of context-free phrase structure rules to describe hierarchical and sequential aspects of linguistic units separately.

Example

Context-free phrase structure rules can be interpreted as well-formedness conditions for local trees. A rule like X $\rightarrow$ $Y_1$ $Y_2$ ... $Y_n$ licences exactly one tree. This tree is characterised by two facts: (i). The root labelled $X$ (directly) dominates the nodes labelled $Y_1$ ... $Y_n$ and (ii). The node labelled $Y_i$ precedes node $Y_{i+1}$ (for i = 1, ... n-1).

An ID-rule X $\rightarrow$ $Y_1,$ $Y_2,$ ..., $Y_n$ on the other hand, licences all trees in which X dominates the other categories, no matter in which order these categories appear. An LP-rule restrains the set of local trees licenced by an ID-rule: Each tree which violates the LP-rule is disregarded.

Comment

Replacing context-free rules by ID- and LP-rules was motivated by the idea that by this division of labor it is possible to state important generalization concerning the word order (constituent order) for a given natural language explicitly.

Each ID/LP-syntax can be converted into a strongly equivalent context-free phrase structure syntax. As the linear restrictions expressed by LP-rules are interpreted globally (no local tree licenced by any ID-rule may violate any of the LP-rules), there are context-free grammars which can not be converted into a strongly equivalent ID/LP-grammar (though into a weakly equivalent ID/LP-syntax).

Other Languages

 REF This article has no reference(s) or source(s). Please remove this block only when the problem is solved.