In comparison to the performance in visual object identification tasks, humans gravely underperform when it comes to naming odors. The poor ability in humans to identify olfactory stimuli has since long been established in psychophysical research; yet, the root cause of this peculiar shortcoming remains essentially unknown. Two primary explanations have been hypothesized: The first posits that poor odor naming is a consequence of neuroanatomical constraints limiting the sensory processing ability of cortical olfactory systems as well as their communication with cortical regions responsible for lexical and semantic representations. In contrast, the second hypothesis proposes that inability to name odors is caused by a mixture of social, cultural, and linguistic factors, whereby humans fail to learn strong and well-defined odor-word associations due to a lack of sufficiently odor-specific lexical labels combined with a negligence of accurate and consistent odor descriptions in everyday written and verbal communication. In this study, we attempt to disentangle and quantify the premise of the latter hypothesis. By applying computational linguistic techniques for semantic content analysis on a corpus of tens of millions of documents published online on a wide variety of topics, we quantify the semantic content, semantic similarity and usage frequency of a set of odor-descriptor words used in a previous psychophysical study to classify odors (Dravnieks, 1985). Crucially, we disambiguate between the semantic content in olfactory and non-olfactory contexts, allowing for an estimation of the semantic ambiguity (number of different meanings attributed to the word), olfactory ambiguity (number of types of smells related to the word), commonness (relative frequency in all contexts), and odor applicability (relative frequency in olfactory contexts) of the odor descriptors. These metrics are compared to the applicability values of the descriptors as reported in Dravnieks’ dataset (1985).