Learning Domain Structures

Abstract

How do people acquire and use knowledge about domain structures, such as the tree-structured taxonomy of folk biology?These structures are typically seen either as consequences of innate domain-speci c knowledge or as epiphenomena of domain-general associative learning.We present an alternative: a framework for statistical inference that discovers the structural principles that best account for di erent domains of objects and their properties.Our approach infers that a tree structure is best for a biological dataset, and a linear structure ("left"-"right") is best for a dataset of people and their political views.We compare our proposal with unstructured associative learning and argue that our structured approach gives the better account of inductive generalization in the domain of folk biology.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 42 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.