Resource-Rational Lossy-Context Surprisal (Model Files)

Resource-Rational Lossy-Context Surprisal is a computationally implemented model of how humans process language, predicting at what points in complex sentences they experience comprehension difficulty. It unifies the memory-based and expectation-based paradigms in psycholinguistics, and provides a more refined account of when hierarchical structure is difficult for humans.

This repository contains both basic parameter files for the inference networks, and a number of replicates of the retention probability parameter files, controlling for variability in different model parameters.

The model is developed by Michael Hahn. It was first described in the following publication:

Lexical Effects in Structural Forgetting: Evidence for Experience-Based Accounts and a Neural Network Model

(Michael Hahn, Richard Futrell, Edward Gibson), 33rd Annual CUNY Human Sentence Processing Conference, 2020

https://people.umass.edu/bwdillon/CUNY_2020_Abstracts/HAHN_Final_Manuscript.pdf