Contextual Resolution of Global Ambiguity for MT

摘要

In this paper, we explore the integration of context into domain-independent machine translation based on a pseudo-semantic approach. We choose a theory of discourse structure (SDRT) to provide the model of context. The incompatibility between the knowledge-poor translation model with the knowledge-rich discourse theory leads to supplement the first with some basic lexical semantics, and to replace the specific rules of the second with more general pragmatic principles. How contextual information is used to choose the preferred interpretation of globally ambiguous sentences is then described with two examples.