=Paper=
{{Paper
|id=Vol-2130/short1
|storemode=property
|title=Emoji Grammar as Beat Gestures
|pdfUrl=https://ceur-ws.org/Vol-2130/short1.pdf
|volume=Vol-2130
|authors=Gretchen McCulloch,Lauren Gawne
}}
==Emoji Grammar as Beat Gestures==
https://ceur-ws.org/Vol-2130/short1.pdf
Emoji Grammar as Beat Gestures
Gretchen McCulloch Lauren Gawne
Lingthusiasm La Trobe University
gretchen.mcculloch@gmail.com l.gawne@latrobe.edu.au
But language isn’t just a list of words. Language
has structures such as subjects and predicates, verbs
Abstract and objects, nouns and adjectives that modify them. If
emoji are truly linguistic, they should also show similar
Emoji are popularly characterized as a “lan- structural properties as words do. In other words, if
guage”, but languages have grammar. What emoji are language, emoji must have a grammar. This
does an emoji grammar look like? Draw- paper searches for a “grammar of emoji” by looking at
ing from sequences of the most common sequences of emoji from a corpus of over 1 billion emoji
two, three, and four emoji in a large cor- uses [McC16] in comparison to the expected sequences
pus of real emoji use, we find that top based on a large corpus of English words [Dav16] and
emoji sequences have a high level of repetition to an alternate hypothesis from the field of gesture
(∼50%), whereas the equivalent top sequences studies, the beat gesture.
of words from a large corpus have zero repeti- Emoji have also been analysed as a strategy for in-
tion. We argue that emoji are best analogized dicating the emotional effect of written speech, which
to “beat” gestures, a well-established type of is usually born by prosody and facial expression in
co-speech gesture characterized by its high spoken language [Miy07] [Wag16]. For example, Face
level of repetition.
With Tears Of Joy can indicate a message is in-
tended to be humorous. While we agree that this is
1 Introduction one important function of emoji, we do not believe that
The use of emoji, small pictures encoded as text this accounts for them fully: many common emoji,
(chiefly faces, handshapes, and common objects), is such as the heart, and all of the objects like food
often characterized as “language” or “linguistic” in pop- and animals, do not have straightforward effects on
ular writing (e.g. [Tho16]). Language is comprised of prosody. Even those emoji that do have emotional or
multiple levels. At a simple approximation, we can say prosodic functions also have lexical correlates: [Dim15]
that a language has phonemes, which combine to cre- found that is used in similar contexts as “lmao,”
ate lexical items (colloquially, words), which combine
while is used like “ugh.” Analyzing the structure
again to create phrases.
of emoji in terms of words is thus not inconsistent with
If emoji correspond to any of these levels, it is that
them having a range of functions, as words do.
of the word. For example, the emoji stands for
the word "heart" or "love", not the /h/ phoneme
1.1 Option 1: Words
or a phrase like “my dog loves pizza.” To convey
“my dog loves pizza” in emoji, one would need, at When analyzing large corpora of language for struc-
minimum, emoji corresponding to “dog,” “love,” and tural recurrences, it is common to analyze them in
“pizza,” again reinforcing that words are the clear level terms of ngrams: recurring sequences of the same n
of comparison. number of words, such as bigrams (2 words), trigrams
(3 words), and quadrigrams (4 words). Perhaps the
Copyright c 2018 held by the author(s). Copying permitted for most well-known tool for analyzing ngrams is Google
private and academic purposes.
Books Ngrams, where one can find that, for exam-
In: S. Wijeratne, E. Kiciman, H. Saggion, A. Sheth (eds.): Pro-
ceedings of the 1st International Workshop on Emoji Under-
ple, the sequence ‘telephone operator’ had a sharp rise
standing and Applications in Social Media (Emoji2018), Stan- in the 1910s, and has been decreasing steadily in fre-
ford, CA, USA, 25-JUN-2018, published at http://ceur-ws.org quency since the 1940s. Here, as we’re looking for a
snapshot of the most frequent ngrams in contempo-
rary English as a whole rather than a historic view of
particular ngrams, we computed the top 200 bigrams,
trigrams, and quadrigrams from COCA, the Corpus of
Contemporary American English [Dav16], which con-
tains around 500 million words from a variety of En-
glish texts such as news outlets and websites.
As expected, there are many overlaps between the
bigram, trigram, and quadrigram lists. For example,
“of the” is a common bigram, while “end of the” is a
common trigram and “the end of the” is a common Figure 1: Most common emoji combinations
quadrigram. Within these top 200 of each, however, .
there are zero instances of purely identical sequences,
i.e. where the same word is repeated to form the en- egories (deixis and emblems, respectively), but either
tire ngram. Such identical sequences are possible in can be produced by moving the hand slightly back and
English (e.g. “had had” and “very very very”), but forth for emphasis, i.e. in the style of a beat. We ar-
they are rare and thus not found in the top 200 lists. gue that repetition of emoji does not have to distract
Drilling down further into the data, we see that when from its other functions (e.g. representing prosodic
the same word appears more than once in a trigram or information), but can co-occur.
quadrigram, it is at the edges of complex constructions
such as “as well as”, “the end of the” and “the rest of 2 The SwiftKey Corpus
the”. To decide between these two options, we look at emoji
While COCA is a corpus of formal English, and ngrams in a corpus we’ll call the SwiftKey Corpus.
emoji are often used in informal contexts, pure rep- This corpus was collected from real-life emoji use by
etition is not common in any variety of English. For users of the SwiftKey smartphone keyboard app on
example, COCA has 585,083 instances of “very” of any both iOS and Android between January 2016 and
kind, of which 442 (0.076%) are “very very” or longer April 2016 who had opted into the use of SwiftKey
[Dav16]. In comparison, the Corpus of Global Web- cloud data for more accurate predictions and had their
Based English (GloWbE) has instances of 14,493 “very language set to US English, containing over a billion
very” or longer versus 2,345,058 “very” of any kind, a instances of emoji use by English speakers. The most
ratio of 0.061% [Dav13]. frequent sequences of emoji were programmatically ex-
tracted from the data as a whole and analyzed as
1.2 Option 2: Beat gestures a list by frequency. So as to preserve user privacy
and anonymity, no individual examples of emoji use
In comparison to grammar, repetition is common in were examined. The SwiftKey Corpus was initially
the gestural domain. There is no such large public cor- created for a talk at South by Southwest by Medlock
pus of gestures for numerical comparison, but a par- and McCulloch [McC16] and subsequently re-analyzed
ticular gesture type, the “beat” gesture, is regularly with additional theoretical framework contributed by
defined as one that contains a repetitive up-down or Gawne for this paper.
side-to-side rhythm [McN92] [McN05]: 40-41; [Ken04]:
103-104, see also [Efr72] ‘baton’; [Fri69] ‘rhythmic’.
The beat gesture is readily observed in both regular
3 Results
conversation, often for emphasizing the rhythm of the The 10 most common sequences of two, three, and
accompanying speech (e.g. one might gesture rhythmi- four emoji (bigrams, trigrams, and quadrigrams) in
cally in a circle while saying, “You just keep going on the SwiftKey Corpus are listed in Table 1 [McC16]; we
and on and on”) and oratory (e.g. a confident speaker analyzed up to the top 200 of each, with and without
might thump rhythmically on a podium to emphasize identical emoji sequences.
their words, while a nervous speaker might jiggle their Validating the emoji ngram approach, there is con-
hands while talking). siderable similarity between the most common emoji
Because a beat refers to the repetitive iteration of on all three lists, similar to what we saw with the over-
a gesture but all gestures must also involve some sort lap between top bigram, trigram, and quadrigram lists
of hand shape in some sort of location, beats read- for the word corpus. However, unlike the word corpus,
ily overlap with other categories of gestures [McN92] it is immediately evident that there is a very high de-
[McN05]: 38, 41. For example, a pointing index finger gree of repetition in the SwiftKey Corpus, which is
and the thumbs up are each classified in other cat- consistent with anecdotal evidence reported for other
emoji datasets. Looking at the top 200 most common towards? We point to the accompanying words.) Fur-
sequences each of two, three, and four emoji in the ther, many emoji appear on the ngrams lists in both
SwiftKey Corpus, roughly half of each are completely orders, something that is very much atypical for words
identical repeats (53%, 52%, and 39.5%), with the pro- in English: “birthday happy” is not the same thing as
portion of non-identical sequences of emoji increasing “happy birthday" and yet both “ ” and “ ”
as one progresses further down each list. are common emoji sequences, or occur on larger strings
The first non-repeating emoji sequences show up of emoji.
at #10 on the bigram list ( ) and #23 on the In contrast, the prototypical use of both emoji and
beat gestures is one of repetition. In fact, the thumbs
trigram ( ) and quadrigram ( )
up emoji directly appears in the top 10 emoji ngrams
lists. Within the non-identical sequences, there re-
lists, just as repeating the thumbs up emblem gesture
mains a high degree of internal repetition. Look-
serves as a beat. Gestures also have the desired flexi-
ing only at the top 200 non-identical trigrams and
quadrigrams, over half contain a partial repetition, bility in terms of sequence ordering: like with
in sequences such as aab, abb, and aba for trigrams and one could equally well point at a person
(75.5%), and aabb, abab, aaab, abbb for quadrigrams and then a cake to ask if the other person wanted some
(67.5%). (Non-identical bigrams were not counted, as cake, or to the cake and then the person for the same
they must consist of ab.) meaning.
Even within entirely heterogeneous sequences (i.e. [McC16] further reports that most (85%) of
abc for trigrams and abcd for quadrigrams), all of SwiftKey sessions containing any emoji do so along-
the top 200 non-identical sequences were thematically side words, and of the sessions containing only emoji,
similar. Such sequences are heterogeneous at a Uni- the majority are only one to two emoji long, presum-
code character encoding level, but not to a human ob- ably a reply to a previous message. This reinforces
server, containing hearts of different colours or shapes another characteristic of the beat gesture, which is its
(such as ), several different monkey faces close relationship with words spoken at the same time,
although further research of a more fine-grained nature
( ), faces of similar emotional valence (such is necessary in order to determine what the details of
as , and related clusters of objects (such that relationship.
as and ). The only sequence
in the top 200 non-identical bigrams, trigrams, and 5 Conclusion
quadrigrams that could possibly depict a scene is When examining sequences of emoji in use, we have
found the most illuminating analysis to be that of
and , but this is more plausibly a depiction of the
emoji as digital gestures, rather than as a grammar
two-handed gesture that it resembles (both used to
with hierarchical structure. In the same way that ges-
represent coitus). No sequences containing simultane-
tures do not have the same grammatical structure as
ously an attitudinal emoji (such as a face or a heart)
speech, but act in concert with it, emoji are not taking
and an object emoji (such as food or birthday items)
on the function of grammar, but acting in relation to
were in the top 200 lists at all.
written text. In particular, repetition of emoji serves
an emphatic function that parallels the use of beat
4 Analysis gestures in spoken discourse.
Repetition is abundant in emoji sequences, and is rare While we have focused on beat gestures in this anal-
in speech or written text. It is not impossible to repeat ysis of emoji sequences, we see many other parallels
identical words in English, such as “very very very” or between the use of gestures with spoken language,
“I love love love love it” for emphasis, and salad-salad and emoji with written language. Other gestural cate-
(in contrast to, say, pasta salad or potato salad) for gories also show promise for understanding the remain-
contrastive focus [Rus04]. Similarly, one could, in prin- der of the emoji paradigm, which we plan to explore
ciple, write heterogeneous sequences of emoji contain- in upcoming work [McC]. In particular, many popu-
lar handshape emoji directly represent the category of
ing subjects and predicates (e.g. to mean emblem gestures, and some extended emoji-only se-
“my dog loves pizza” or to mean “I am happy quences parallel the gesture category of pantomime
when I drink beer”). However, neither is a prototypi- (see Emoji Dick, [Ben10] for one of the most elabo-
cal use, as no such attitude/object pairings are found rate manifestations of “emoji pantomime”).
on the top 200 ngram lists. (In this case, one might It is, perhaps, unsurprising that people use emoji
ask, what is a happy face emoji indicating an attitude in digital communication in ways that parallel use of
co-speech gesture, given that gesture has important eastern Jews and southern Italians in New
functions both for communication [Hey75] [Coe18] and York City, living under similar as well as
cognition [GM98] [Chu17]. Treating emoji as gesture different environmental conditions. Mouton,
makes it clear that emoji are unlikely to become a lan- 1941/1972.
guage in their own right. Languages that draw on the
same modality as gestures are Signed Languages, and [Fri69] P. Ekman & W.V. Friesen. The repertiore
have structural properties that are are more similar to of nonverbal behaviour: categories, origins,
spoken languages than to co-speech gesture, i.e. pre- usage, and coding. Semiotica, 1:49–98, 1969.
cisely the structural regularities that we have demon- [GM98] J. Iverson & S. Goldin-Meadow. Why
strated that emoji do not currently have. If emoji do people gesture when they speak. Nature,
ever emerge as a language proper, we will find it by 396(6708):228, 1998.
seeing these same structural regularities emerge in a
large corpus study like the one in this paper. [Hey75] J.A. Graham & S. Heywood. The effects of
elimination of hand gestures and of verbal
5.0.1 Acknowledgements codability on speech performance. European
Gretchen McCulloch would like to thank the team at Journal of Social Psychology, 5(2):189–195,
SwiftKey, including Jennifer Kutz and Nicky Budd- 1975.
Thanos, for running data queries and facilitating the [Ken04] A. Kendon. Gesture: Visible action as utter-
talk at South by Southwest. Lauren Gawne would like ance. Cambridge University Press, 2004.
to thank La Trobe University and the David Myers
Fellowship Program. Both authors would like to thank [McC] L. Gawne & G. McCulloch. Emoji are digital
three anonymous reviewers for their helpful comments. gesture (in prep).
[McC16] B. Medlock & G. McCulloch. The linguistic
References secrets found in billions of emoji. Presented
[Ben10] F. Benenson. Emoji Dick. Lulu, 2010. at SXSW, March 11-20, Austin, Texas, 2016.
[Chu17] S. Kita M. Alibali & M. Chu. How do [McN92] D. McNeill. Hand and mind: What ges-
gestures influence thinking and speaking? tures reveal about thought. The University
the gesture-for-conceptualization hypothesis. of Chicago Press, 1992.
Psychological review, 124(3):245, 2017.
[McN05] D. McNeill. Gesture and thought. University
[Coe18] T. Jenkins M. Coppola & C. Coelho. Effects of Chicago Press, 2005.
of gesture restriction on quality of narrative
production. Gesture, 16(3):416–431, 2018. [Miy07] K. Miyake. How young Japanese express
their emotions visually in mobile phone mes-
[Dav13] M. Davies. Corpus of Global Web- sages: A sociolinguistic analysis. Japanese
Based English: 1.9 billion words from Studies, 27(1):53–72, 2007.
speakers in 20 countries (GloWbE).
https://corpus.byu.edu/glowbe/, 2013. [Rus04] J. Ghomeshi R. Jackendoff N. Rosen & K.
Russell. Contrastive focus reduplication in
[Dav16] M. Davies. The Corpus of Contempo- English (the salad-salad paper). Natural
rary American English (COCA): 560 mil- Language & Linguistic Theory, 22(2):307–
lion words, 1990-present. Available online at 357, 2004.
https://corpus.byu.edu/coca/, 2008-2016.
[Tho16] C. Thompson. The emoji is the birth of
[Dim15] T. Dimson. Emojineering part 1: Ma- a new type of language (no joke). Wired,
chine learning for emoji trends. Insta- 24.04.16, 2016.
gram Engineering blog https://instagram-
[Wag16] M. Wagner. How to be kind with prosody.
engineering.com/emojineering-part-1-
Speech Prosody 2016, pages 1250–1253, 2016.
machine-learning-for-emoji-trendsmachine-
learning-for-emoji-trends-7f5f9cb979ad, May
1, 2015.
[Efr72] D. Efron. Gesture, race and culture; a ten-
tative study of the spatio-temporal and ’lin-
guistic’ aspects of the gestural behavior of