Computational Modeling of an Auditory Lexical Decision Experiment Using DIANA.

Area of Science:

• Cognitive Science
• Computational Linguistics
• Psycholinguistics

Background:

• Spoken word recognition is a complex cognitive process involving rapid identification of words from continuous acoustic signals.
• Computational models aim to simulate and understand the mechanisms underlying human spoken word recognition.
• The Massive Auditory Lexical Decision (MALD) project provides valuable behavioral data for testing such models.

Purpose of the Study:

• To implement and evaluate DIANA, an end-to-end computational model of spoken word recognition.
• To assess DIANA's performance on lexical decision tasks using data from the MALD project.
• To investigate DIANA's ability to model human response latencies in spoken word recognition.

Main Methods:

• Development of acoustic models for DIANA to process novel speech input.
• Simulation of DIANA's performance in distinguishing words from pseudowords.
• Generation and correlation of estimated response latencies with MALD participant data.

Main Results:

• DIANA demonstrated effective performance in free word recognition and lexical decision tasks.
• The model's estimations of response latency were found to be inversely correlated with observed behavioral data.
• Discrepancies highlight limitations in the current latency estimation approach within DIANA.

Conclusions:

• DIANA serves as a viable computational tool for modeling aspects of spoken word recognition.
• Further refinement of latency prediction mechanisms is necessary for DIANA to fully capture human behavioral patterns.
• The study provides insights into the requirements for contemporary spoken word recognition models.