DIANA, a Process-Oriented Model of Human Auditory Word Recognition.

Area of Science:

• Cognitive Science
• Neuroscience
• Computational Linguistics

Background:

• Existing models of auditory word recognition often lack detailed neuro-physiological grounding.
• Human speech processing involves complex interpretation of acoustic signals and task-dependent decision-making.

Purpose of the Study:

• To introduce DIANA, a novel process-oriented computational model for human auditory word recognition.
• To enhance psycholinguistic research by providing a model that outputs word identifications, lexicality decisions, and reaction times comparable to human listeners.

Main Methods:

• DIANA models auditory word recognition by taking the acoustic signal as input.
• It incorporates neuro-physiological evidence, including spectro-temporal receptive fields found in the human superior temporal gyrus.
• The model comprises activation, decision, and execution components, with detailed conceptual and computational descriptions.

Main Results:

• DIANA accounts for the impact of acoustic signal ambiguity on reaction times, aligning with the Hick-Hyman law.
• It interprets acoustic signals using spectro-temporal receptive fields rather than abstract phonological units.
• The model's activation component is task-independent, while its decision component is task-dependent.

Conclusions:

• DIANA offers a more neuro-physiologically informed approach to modeling human auditory word recognition.
• The model's ability to generate comparable outputs to human listeners facilitates psycholinguistic experimentation.
• Future improvements aim to further enhance DIANA's resemblance to human listener behavior.