The semantic representation of locatives in machine translation
Fredrik Jørgensen · NORA - Norwegian Open Research Archives · 2004
(1) Musa løp under bordet. 'The mouse ran under the table.' (2) Under bordet er et fint sted å gjemme seg. 'Under the table is a good place to hide.' Locative prepositional phrases, as seen in (1) and (2) ('under bordet'), all express spatial concepts, but they can be interpreted in a number of different ways. In (1), the mouse may (i) be running around under the table, (ii) run to the region under the table (e.g. to hide), or (iii) run through the region under the table (e.g. 'Musa løp under bordet og ut på den andre siden'). These three different readings all modify the event or situation of the mouse's running, but differ in the way they modify it. In (2), however, the prepositional phrase seems to refer to a region of space rather than modifying an event. I describe a theory on how to interpret locative prepositional phrases, which tries to account for both the modificational and the referential proeprties of preositional phrases, nameley the theory presented by Marcus Kracht in 'On the Semantics of Locatives' (Linguistics and Philosophy, 25:157-232, 2002). Furthermore, I implement the main features of this theory in a computational framework, the 'Linguistic Knowledge Builder' system (LKB), using 'Head-driven Phrase-Structure Grammar' (HPSG) as the theory for syntactic representation and 'Minimal Recursion Semantics' (MRS) as the theory for semantic representation. I develop a model (grammar fragment) for Norwegian locatives, and explore how this model describes Norwegian locatives with respect to their syntactic behaviour and semantics properties. I argue that the theory of Kracht (2002) predict how prepositional phrases with different syntactic structures express the same type of semantics (meaning). I also apply the theory of Kracht (2002) in a contrastive analysis of locative prepositional phrases in Norwegian and English, and argue that we can use this model for more accurate translations in a semantic transfer-based Machine Translation system.