Shifting meaning representations
Karen Spärck Jones · International Joint Conference on Artificial Intelligence · 1983
This paper argues that a number of different kinds of meaning representation, between which partial translations can be made as needed, are all required for a reasonably comprehensive language processing system. These representations capture different and possibly complementary aspects of a text's form, content and reference worlds, and are suited to different subtasks of the language processor. Initial testing of the proposition via a system designed for natural language access to databases is described, showing how different types of meaning representation with different characteristics are called for, related, and used.