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

Abstract

This article presents DIANA, a new, process-oriented model of human auditory word recognition, which takes as its input the acoustic signal and can produce as its output word identifications and lexicality decisions, as well as reaction times. This makes it possible to compare its output with human listeners’ behavior in psycholinguistic experiments. DIANA differs from existing models in that it takes more available neuro-physiological evidence on speech processing into account. For instance, DIANA accounts for the effect of ambiguity in the acoustic signal on reaction times following the Hick–Hyman law and it interprets the acoustic signal in the form of spectro-temporal receptive fields, which are attested in the human superior temporal gyrus, instead of in the form of abstract phonological units. The model consists of three components: activation, decision and execution. The activation and decision components are described in detail, both at the conceptual level (in the running text) and at the computational level (in the Appendices). While the activation component is independent of the listener’s task, the functioning of the decision component depends on this task. The article also describes how DIANA could be improved in the future in order to even better resemble the behavior of human listeners.

Semantic richness effects in isolated spoken word recognition: Evidence from massive auditory lexical decision.

“…Still, such a representation of the mental lexicon does not capture effects related to rating-based measures of semantic richness like concreteness, valence, or arousal. It seems that models that make use of top-down weights, such as TRACE (McClelland & Elman, 1986) or DIANA (Ten Bosch et al, 2015, 2022), have a way of capturing these effects if the values of these variables are stored within the lexicon. It would be best if we could represent these through separate top-down weights or resting activation increases in order to avoid conflating different factors under a single factor.…”

“…Instead, the process of lexical access in models of spoken word recognition most often begins with the processing of (pseudo)acoustic features or phonemes and ends with detection of the correct form (unit/word) in the mental lexicon, while effectively eschewing any meaning activation (see also Gaskell & Marslen-Wilson, 2002, for additional critique). The mental lexicon is then usually represented as a semantically unconnected list of words—words are often strings of phonemes, related to each other by form only (as in, e.g., Hannagan et al, 2013; Luce, 1986; Luce et al, 2000; Luce & Pisoni, 1998; Marslen-Wilson, 1987; Marslen-Wilson & Tyler, 1980; McClelland & Elman, 1986; Norris, 1994; Norris & McQueen, 2008; Ten Bosch et al, 2015; You & Magnuson, 2018).…”

“…However, certain models can potentially provide a means to explain and simulate the role of semantics in the process of spoken word recognition with minor modification. For example, TRACE (McClelland & Elman, 1986) and DIANA (Ten Bosch et al, 2015, 2022) both posit a feedback mechanism designed to account for word frequency effects. Such a mechanism could be employed to account for the effects of other variables as well, including those related to a word’s semantic richness.…”

Processing reduced speech in the L1 and L2: a combined eye-tracking and ERP study

“…Many theories of spoken word recognition assume the existence of some kind of mental lexicon that links phonetic representations of words with their syntactic and semantic features (see e.g. Cutler, 2012 ; ten Bosch et al, 2022 and the citations therein for more details). Theories agree that the activation of words' representations involves both bottom-up (acoustic-phonetic matching) and top-down (context-depending prediction).…”

Modelling Human Word Learning and Recognition Using Visually Grounded Speech

“…The exact set of units is predetermined by the model developer, avoiding the issue of learning what these units are in the Multimedia Computing Group, Delft University of Technology, Delft, The Netherlands first place. Even the recently introduced DIANA model [10], which does away with fixed pre-lexical units, uses a set of predetermined lexical units.…”