This paper focuses on the semantic representation of verbs in computer systems and its impact on lexical selection problems in machine translation (MT). The authors examine two groups of English and Chinese verbs to demonstrate that lexical selection must be based on sentence interpretation and verb argument selection restrictions. They propose a novel verb semantic representation scheme that defines each verb by a set of concepts in different conceptual domains, allowing for correct lexical selection even when exact matches are not available. This approach is compared to transfer-based MT methods and is seen as compatible with knowledge-based MT (KBMT) approaches. The paper highlights the limitations of direct transfer methods, particularly in handling idiomatic usages and complex verb usages. It introduces a multi-domain approach that uses a verb taxonomy to relate verbs with similar meanings and defines verb projections based on shared semantic domains. The authors implement a prototype system, UNICON, which uses this representation to achieve accurate lexical selection. Experimental results show that the proposed method improves the accuracy of lexical selection, especially in handling metaphorical usages. The paper concludes by discussing the benefits of the proposed method and the potential for future work in scaling up the system.This paper focuses on the semantic representation of verbs in computer systems and its impact on lexical selection problems in machine translation (MT). The authors examine two groups of English and Chinese verbs to demonstrate that lexical selection must be based on sentence interpretation and verb argument selection restrictions. They propose a novel verb semantic representation scheme that defines each verb by a set of concepts in different conceptual domains, allowing for correct lexical selection even when exact matches are not available. This approach is compared to transfer-based MT methods and is seen as compatible with knowledge-based MT (KBMT) approaches. The paper highlights the limitations of direct transfer methods, particularly in handling idiomatic usages and complex verb usages. It introduces a multi-domain approach that uses a verb taxonomy to relate verbs with similar meanings and defines verb projections based on shared semantic domains. The authors implement a prototype system, UNICON, which uses this representation to achieve accurate lexical selection. Experimental results show that the proposed method improves the accuracy of lexical selection, especially in handling metaphorical usages. The paper concludes by discussing the benefits of the proposed method and the potential for future work in scaling up the system.