A semantic analysis of parosmia identifying underlying semantic factors and determining parosmia severity on the basis of natural language data
Fri-P2-088
Presented by: Thomas Hörberg
Parosmia is an olfactory disorder that involves distortions of specific odors (e.g., experiencing the smell of freshly brewed coffee as rotten fruit). Parosmia for some odors may co-occur with anosmia for other odors. Little is known about which odors are most frequently parosmia triggers, and tools for determining parosmia severity are lacking.
We present a new approach to understand and diagnose parosmia that is based on semantic properties of words associated with specific odors. Using a new data-driven method based on natural language data (Hörberg et al. 2020; Hörberg et al. under review), we identified 38 source-based olfactory descriptors (e.g., garlic) that are evenly dispersed across a large olfactory-semantic space. Parosmia patients (n = 48) classified these 38 descriptors in terms of whether their corresponding odors were perceptually distorted (i.e., parosmic), or whether they were completely odorless (i.e., anosmic). We then investigated whether these parosmic and anosmic classifications are related to lexical-semantic properties of the original words. We found parosmic experiences to most commonly be reported for words denoting unpleasant odors of inedibles that are highly associated to olfaction (e.g., excrement). Anosmic experiences, on the other hand, were most frequently reported for words that describe highly specific odor experiences of edibles but that are less strongly associated to olfaction (e.g., pistachio) in comparison to other descriptors (e.g., perfume).
Based on PCA modeling of our data, we finally derived the Parosmia Severity Index-a measure of the severity of parosmia that can be determined solely on the word classification task. We found this index to be predictive of olfactory sensitivity, self-reported olfactory impairment, and depression. Our work provides a novel approach for investigating parosmia and a novel tool for establishing parosmia severity that does not require perceptual screening with odors.
We present a new approach to understand and diagnose parosmia that is based on semantic properties of words associated with specific odors. Using a new data-driven method based on natural language data (Hörberg et al. 2020; Hörberg et al. under review), we identified 38 source-based olfactory descriptors (e.g., garlic) that are evenly dispersed across a large olfactory-semantic space. Parosmia patients (n = 48) classified these 38 descriptors in terms of whether their corresponding odors were perceptually distorted (i.e., parosmic), or whether they were completely odorless (i.e., anosmic). We then investigated whether these parosmic and anosmic classifications are related to lexical-semantic properties of the original words. We found parosmic experiences to most commonly be reported for words denoting unpleasant odors of inedibles that are highly associated to olfaction (e.g., excrement). Anosmic experiences, on the other hand, were most frequently reported for words that describe highly specific odor experiences of edibles but that are less strongly associated to olfaction (e.g., pistachio) in comparison to other descriptors (e.g., perfume).
Based on PCA modeling of our data, we finally derived the Parosmia Severity Index-a measure of the severity of parosmia that can be determined solely on the word classification task. We found this index to be predictive of olfactory sensitivity, self-reported olfactory impairment, and depression. Our work provides a novel approach for investigating parosmia and a novel tool for establishing parosmia severity that does not require perceptual screening with odors.