Soo Hyun Ryu
Biographic Data
| ID | 6090657 |
|---|---|
| NAME | Soo Hyun Ryu |
| GIVEN NAMES | Soo Hyun |
| FAMILY NAME | Ryu |
| SIGNATURE | RYU S H |
| AFFILIATIONS | University of Michigan |
| ORCID | 0000-0001-6636-9765 |
| VERIFIED | Yes |
| TOTAL WORKS | 2 |
| TOTAL CITATIONS | 0 |
| WORKS AS AUTHOR | 2 |
| WORKS AS EDITOR | 0 |
| FIRST PUBLICATION YEAR | 2022 |
| LATEST PUBLICATION YEAR | 2025 |
| H-INDEX | 0 |
Memory for prediction
We demonstrate that Transformer-based neural network language models provide a new foundation for mechanistic theories of sentence processing that seamlessly integrate expectation-based and memory-based accounts. First, we show that the attention mechanism in GPT2-small operates as a kind of cue-based retrieval architecture that is subject to similarity-based interference. Second, we show that it provides accounts of classic memory effects in par…
Flexible Acceptance Condition of Generics from a Probabilistic Viewpoint
Formalization of the semantics of generics has been considered extremely challenging for their inherent vagueness and context-dependence that hinder a single fixed truth condition. The present study suggests a way to formalize the semantics of generics by constructing flexible acceptance conditions with comparative probabilities. Findings from our in-depth psycholinguistic experiment show that two comparative probabilities—cue validity and preval…
No prominent works on this page.
Flexible Acceptance Condition of Generics from a Probabilistic Viewpoint
Formalization of the semantics of generics has been considered extremely challenging for their inherent vagueness and context-dependence that hinder a single fixed truth condition. The present study suggests a way to formalize the semantics of generics by constructing flexible acceptance conditions with comparative probabilities. Findings from our in-depth psycholinguistic experiment show that two comparative probabilities—cue validity and preval…
Memory for prediction
We demonstrate that Transformer-based neural network language models provide a new foundation for mechanistic theories of sentence processing that seamlessly integrate expectation-based and memory-based accounts. First, we show that the attention mechanism in GPT2-small operates as a kind of cue-based retrieval architecture that is subject to similarity-based interference. Second, we show that it provides accounts of classic memory effects in par…
Cognitive psychology (2 works) · Computer Science (2 works) · Natural language processing (2 works) · Psychology (2 works) · Advanced Text Analysis Techniques (1 work) · Artificial Intelligence (1 work) · Artificial Intelligence (1 work) · Biomedical Text Mining and Ontologies (1 work) · Cognition (1 work) · Construct (python library) (1 work)