Verb Semantics and Lexical Selection

The paper focuses on the semantic representation of verbs in computer systems and its impact on the lexical selection problem in machine translation. Using two groups of English and Chinese verbs as evidence, it argues that choosing the correct target-language verb cannot rely on selection restrictions over verb arguments alone but must also incorporate an interpretation of the sentence. The authors introduce a novel representation scheme and contrast it with the selection-restriction representations typically used in transfer-based machine translation.

They frame their approach as closely aligned with knowledge-based machine translation and as a separate component that could be incorporated into existing systems. Through examples and experimental results, they show that with this scheme even inexact matches can still produce correct lexical selection, indicating that richer verb semantics can make translation choices more robust.