Controlling the retrieval of general vs specific semantic knowledge in the instance theory of semantic memory
- Matthew Crump, Psychology, Brooklyn College of CUNY, Brooklyn, New York, United States
- Randall Jamieson, Department of Psychology, University of Manitoba, Winnipeg, Manitoba, Canada
- Brendan Johns, Psychology, McGill University, Montreal, Quebec, Canada
- Michael N. Jones, Department of Psychological and Brain Sciences, Indiana University, Bloomington, Indiana, United States
AbstractDistributional models of semantic cognition commonly make simplifying assumptions, such as representing word co-occurrence structure by prototype-like high-dimensional semantic vectors, and limit how retrieval processes may contribute to the construction and use of semantic knowledge. More recently, the instance theory of semantics (ITS,Jamieson, Avery, Johns, and Jones, 2018) reconceived a distributional model in terms of instance-based memory, allowing context-specific construction of semantic knowledge at the time of retrieval. By simulation, we show that additional encoding and retrieval operations, consistent with learning and memory theory, can play a crucial role in flexibly controlling the construction of general versus specific semantic knowledge. We argue this consolidation of processing principles holds insight for distributional theories of semantic cognition.