The language of smell: Connecting linguistic and psychophysical properties of odor descriptors

Abstract

The olfactory sense is a particularly challenging domain for cognitive science investigations of perception, memory, and language. Although many studies show that odors often are difficult to describe verbally, little is known about the associations between olfactory percepts and the words that describe them. Quantitative models of how odor experiences are described in natural language are therefore needed to understand how odors are perceived and communicated. In this study, we develop a computational method to characterize the olfaction-related semantic content of words in a large text corpus of internet sites in English. We introduce two new metrics: olfactory association index (OAI, how strongly a word is associated with olfaction) and olfactory specificity index (OSI, how specific a word is in its description of odors). We validate the OAI and OSI metrics using psychophysical datasets by showing that terms with high OAI have high ratings of perceived olfactory association and are used to describe highly familiar odors. In contrast, terms with high OSI have high inter-individual consistency in how they are applied to odors. Finally, we analyze Dravnieks's (1985) dataset of odor ratings in terms of OAI and OSI. This analysis reveals that terms that are used broadly (applied often but with moderate ratings) tend to be olfaction-unrelated and abstract (e.g., "heavy" or "light"; low OAI and low OSI) while descriptors that are used selectively (applied seldom but with high ratings) tend to be olfaction-related (e.g., "vanilla" or "licorice"; high OAI). Thus, OAI and OSI provide behaviorally meaningful information about olfactory language. These statistical tools are useful for future studies of olfactory perception and cognition, and might help integrate research on odor perception, neuroimaging, and corpus-based linguistic models of semantic organization.

Mapping the semantic organization of the English odor vocabulary using natural language data

“…In this paper, we develop and test a method that uses structured collections of natural texts and machine learning algorithms to automatically identify odor and flavor descriptors and derive the semantic organization of those descriptors. Our method automatically identifies descriptors in texts on the basis of their olfactory and gustatory association (i.e., OAI, see Iatropoulos et al, 2018), as based on their distributions in olfactory and gustatory contexts in natural text collections. The semantic space of these descriptors is then derived using a distributional-semantic word embedding model, which represents semantic distances between words as vector distances in a multi-dimensional space.…”

“…Croijmans et al (2019) used a textbased, computational approach on a collection of wine reviews in order to identify the vocabulary used to describe wine qualities. Iatropoulos et al (2018) developed the method employed in the present study, the Olfactory Association Index (OAI), which quantifies words with respect to how strongly associated they are with olfactory and gustatory contexts. Other studies have used pre-trained distributional-semantic word embedding models to predict the applicability of odor descriptors to odor molecules (Gutiérrez et al, 2018;Nozaki & Nakamoto, 2018).…”

“…They identified 146 unique terms that were used to construct a textbased wine wheel. Iatropoulos et al (2018) proposed a corpus-based measure of olfactory / gustatory association (similar to Termhood analysis), the Olfactory Association Index (OAI). OAI is the log2 probability that descriptor d occurs in a olfactory-related context as opposed to a non-olfactory context, i.e.,…”

“…OAI was found to be strongly correlated (ρ = .693) with experimental ratings of olfactory association, collected by Lynott & Connell (2009), attesting to the viability of OAI as a measure of olfactory association. Iatropoulos et al (2018) further devised a measure of olfactory / gustatory specificity, the Olfactory Specificity Index (OSI), defined as the negative log2-probability that descriptor d occurs in a olfactoryrelated context, where D is the set of all descriptors investigated. As such, OSI estimates the information contained in a descriptor about olfactory contexts.…”

Predicting natural language descriptions of mono-molecular odorants

“…Indeed ratings of a molecule along a comprehensive set of descriptors such as “putrid”, “floral”, and “apple” could uniquely characterize the molecule’s odor 1 , and experts spend considerable time and effort handcrafting domain-specific sets of odor descriptors or collecting ratings for large numbers of descriptors for each molecule of interest 3 , 4 . A standard, generally applicable set of “primary” odor descriptors would be more amenable 5 but despite decades of research this effort has been in vain 2 , 6 . The prevailing view is that there is a significant disconnect between humans’ strong capacity for odor discrimination 7 and their inability to identify or characterize odors by name 1 , 8 – 12 .…”