On the Context-Free Ambiguity of Emoji

Abstract

Due to their pictographic nature, emojis come with baked-in, grounded semantics. Although this makes emojis promising candidates for new forms of more accessible communication, it is still unknown to what degree humans agree on the inherent meaning of emojis when encountering them outside of concrete textual contexts. To bridge this gap, we collected a crowdsourced dataset (made publicly available) of one-word descriptions for 1,289 emojis presented to participants with no surrounding text. The emojis and their interpretations were then examined for ambiguity. We find that, with 30 annotations per emoji, 16 emojis (1.2%) are completely unambiguous, whereas 55 emojis (4.3%) are so ambiguous that the variation in their descriptions is as high as that in randomly chosen descriptions. Most emojis lie between these two extremes. Furthermore, investigating the ambiguity of different types of emojis, we find that emojis representing symbols from established, yet not cross-culturally familiar code books (e.g., zodiac signs, Chinese characters) are most ambiguous. We conclude by discussing design implications.

Translating Polysemous Emojis into Kurdish: An Empirical Study

“…While KevinMcKenzie claims that he was the first who proposed the smiley in an email on April 12, 1979 (Raymond, 1996). For Crystal (2001), emoticons are combinations of keyboard characters invented to show an emotional facial expression. Nearly all of the emoticons are read sideways.…”

“…Therefore, they created a new dataset called EmoSim508, which assigns human annotated semantic similarity scores to a set of 508 carefully selected emoji pairs. Częstochowska et al (2022) found out that emojis come with very different amounts of prepacked semantics. A number of emojis are completely unambiguous, whereas the majority of annotators describe them with the same word.…”

The Lexicon of Emoji? Conventionality Modulates Processing of Emoji

“…without much potential for ambiguity nor room for interpretation. Cz ęstochowska et al (2022) analyzed ambiguity rates for different categories of emoji based on participant descriptions of individual emoji presented without surrounding context. They found "food & drink," "clothes & accessories," "nature," and "hearts" to be the categories with the least variation.…”

“…We first turn to our analysis of offline emoji lexicalization: do people agree on the meanings of emoji? Intuitively and evidently (Częstochowska et al., 2022), some emoji should have high meaning agreement (e.g., many of the animals, foods, and objects). Others, however, are not as straightforward—some are polysemous with multiple widely recognized meanings, like the notoriously euphemistic peach and eggplant, and others may not have any clearly accepted and widely agreed upon meaning.…”

“…Though the dynamics are not quite as clear with emoji as they are with words, examples of both homonymy and polysemy exist in the emoji data collected by Częstochowska et al. (2022). The most frequent descriptions provided for were “evil” and “devil,” and the most frequent for were “sad” and “frown.” These examples both seem to constitute classic examples of polysemy, with the latter acting as an embodiment or a manifestation of the former.…”