Determine the lexical knowledge required for analogical compound interpretation

Determine whether a vocabulary of approximately 33,800 Mandarin words overestimates the lexical knowledge of many educated speakers and thereby whether analogical compound interpretation can operate with substantially smaller individual vocabularies.

Background

The Compound Analogy Model (CAM) relies on local compound families to estimate constituent-specific semantic shifts. For compounds with more than two constituents, these families can be very small, which limits the basis for analogical generalization. The authors note that expanding the lexical database can mitigate this sparsity, but question whether the resulting vocabulary is cognitively realistic for individual speakers.

The authors’ dataset contains approximately 33,800 words, whereas the Modern Chinese Dictionary contains approximately 65,000–70,000 commonly used lexical entries. The unresolved issue is how much lexical knowledge individual speakers actually possess and whether the observed analogical performance depends on an unrealistically large vocabulary.