Evaluating Computational Metrics for Predicting N400 Amplitude during Reading Comprehension
Abstract: Given the interest recent research showed towards cognitive modeling of ERPs, we explored whether traditional word-level features such as position, word frequency, and number of strokes overlap with probability-based metrics such as surprisal, entropy, and entropy reduction. Analyzing and comparing different generalized linear models we found that the mathematical metrics do represent the same information as some of the "traditional" features overpowering them. A new cognitive-motivated computational feature is proposed.
Paper Type: Short
Research Area: Linguistic theories, Cognitive Modeling and Psycholinguistics
Research Area Keywords: Linguistic theories, Cognitive Modeling and Psycholinguistics
Contribution Types: Data analysis, Theory
Languages Studied: Mandarin Chinese
Submission Number: 2120
Loading