Turkish has considerably freer word order than English. The interpretations of different word orders in Turkish rely on information that describes how a sentence relates to its discourse context. To capture the syntactic features of a free word order language, I present an adaptation of Combinatory Categorial Grammars called {}-CCGs (set-CCGs). In {}-CCGs, a verb's subcategorization requirements are relaxed so that it requires a set of arguments without specifying their linear order. I integrate a level of information structure, representing pragmatic functions such as topic and focus, with {}-CCGs to allow certain pragmatic distinctions in meaning to influence the word order of a sentence in a compositional way. Finally, I discuss how this strategy is used within an implemented generation system which produces Turkish sentences with context-appropriate word orders in a simple database query task.