Content uploaded by Yiqiong Zhang
Author content
All content in this area was uploaded by Yiqiong Zhang on Apr 05, 2023
Content may be subject to copyright.
Thank you for your interest in our article on graphicon evolution.
Yiqiong Zhang, Susan Herring, and Suifu Gan. 2022. Graphicon Evolution on the Chinese
Social Media Platform BiliBili. In Proceedings of the Fifth International Workshop on
Emoji Understanding and Applications in Social Media, pages 75–85, Seattle,
Washington, USA. Association for Computational Linguistics.
This is an updated version of our article published by Association for Computational
Linguistics. https://aclanthology.org/2022.emoji-1.9/
After the article was published, we found part of the data from the year of 2021 were
repetitive. In this updated version, we have deleted the repeated messages and corrected the
statistics accordingly. The overall trends and our arguments remain the same as they were
based on normalized frequency. For those who would like to refer to details, please use the
numbers from this corrected version. Thank you.
Graphicon Evolution on the Chinese Social Media Platform BiliBili
Yiqiong Zhang Susan C. Herring Suifu Gan
Guangdong University of Indiana University Bloomington Jinan University
Foreign Studies, China USA China
zhangyiqiong@gdufs.edu.cn herring@indiana.edu 657742829@qq.com
Abstract
This study examines the evolutionary
trajectory of graphicons in a 13-year corpus of
comments from BiliBili, a popular Chinese
video-sharing platform. Findings show that
emoticons (kaomoji) rose and fell in
frequency, while emojis and stickers are both
presently on the rise. Graphicon distributions
differ in comments and replies to comments.
There is also a strong correlation between the
types of graphicons used in comments and
their corresponding replies, suggesting a
priming effect. Finally, qualitative analysis of
the 10 most-frequent kaomojis, emojis, and
stickers reveals a trend for each successive
graphicon type to become less about emotion
expression and more integrated with platform-
specific culture and the Chinese language.
These findings lend partial support to claims
in the literature about graphicon evolution.
1 Introduction
Graphicons are graphical icons used in text-based
computer-mediated communication (Herring &
Dainas, 2017). From the first use of :-) in 1982
(Evans, 2017) to the varied and colorful stickers
on social media today, graphicons have changed
dramatically. ASCII emoticons, the first
graphicons, were composed of keyboard symbols
and were typically used for expressing emotion.
Emoticons in the Western context emphasize the
mouth and are read at a 90-degree angle to the
words (e.g., :-) for a smiley), while kaomoji
(literally ‘face letters’), a style of emoticon that
arose in Japan and also became popular in China,
are read in-line with words and emphasize the
eyes (e.g., ^_^ or ^^) (Katsuno & Yano, 2002).
Kaomojis express not only emotions, but also
actions, objects, and story lines.
1
1
http://kaomoji.ru/en/, retrieved April 5, 2022.
Emojis were adopted globally after Apple
included them in the iPhone in 2010 (Danesi,
2016). Emoji are more colorful, more
representational (as opposed to schematic), and
express a wider array of concepts than ASCII
emoticons. Stickers, which were introduced a few
years after emojis, take these trends further
(Konrad et al., 2020). Usually larger than
emoticons and emojis, stickers may include text;
this is typical of stickers used on Chinese social
media (e.g., see the examples in de Seta, 2018; Ge,
2020). Stickers are character-driven illustrations
or animations that are typically offered as
thematic sets on social media platforms (de Seta,
2018), although social media users in China may
also create their own stickers (Ge, 2020).
Extensive studies have addressed the meaning,
function, and usage of each type of graphicon in
different cultural contexts (e.g., Al Rashdi, 2018;
Ge, 2020; Ge & Herring, 2018; Logi &
Zappavigna, 2021; Sampietro, 2019).
Interrelations among the three types, however,
have not attracted much attention until recently.
Studies have explored the uses of the three
graphicon types (de Seta, 2018), user perceptions
of the three types (Tang & Hew, 2018), and the
evolutionary trends they follow (Konrad et al.,
2020). While these studies provide rich insights,
the first two mainly used qualitative methods, and
the latter analyzed contemporary data, despite
making diachronic claims. Their findings remain
to be verified by empirical comparison of
graphicon use in longitudinal data.
2 Background
2.1 Graphicon evolution
As graphicons continue to grow in popularity
worldwide and shape social media and mobile
communications, it is important to understand
how and why they evolve, and the implications of
their evolution for where they are headed in the
future. Konrad et al. (2020) posit that graphicons
tend to follow an evolutionary trajectory
consisting of three phases: an early phase, a high
(or peak) phase, and a phase of decline and/or
conventionalization. One of the main criteria for
determining which phase a graphicon is currently
in is frequency of use; another is pragmatic
changes in graphicon use. For Western graphicons,
according to Konrad et al. (2020), emoticons are
in the third phase, emoji are in the second phase,
and stickers are in the first phase. The authors
predict that emoji will eventually reach the third
phase, following the path of emoticons, and that
stickers may eventually reach the second (and
eventually the third) phase and overtake emoji in
popularity.
The history of ASCII emoticons and emoji
provide evidence in partial support of this
trajectory. Pavalanathan and Eisenstein (2016)
analyzed emoticons and emoji on Twitter in the
17 months after emoji were first introduced on the
platform. They found that emoticon use
dramatically decreased as emoji use increased.
Furthermore, a number of studies have reported
that emoticons have become conventionalized as
a type of punctuation (Markman & Oshima, 2007;
Provine et al., 2007). That is, emoticons have
declined in frequency of use and have become
conventionalized, evidence that they are in the
third phase of Konrad et al.’s (2020) evolutionary
trajectory. Meanwhile, emojis in the West remain
at the peak of their popularity.
While this evidence is compelling, it is limited.
As yet no comparable evidence exists for all three
graphicon types, or for the relationship of emojis
to stickers. Konrad et al. (2020) interviewed and
surveyed Facebook Messenger users about their
use of emoji and stickers, identifying many areas
of overlap in function of the two graphicons. They
also noted some differences: participants
described emojis as better suited for expressing
emotion, whereas stickers were considered more
specific and better at expressing the user’s
personality. However, Konrad et al. (2020) did not
quantify emoji and sticker use over time. What is
needed is a longitudinal corpus of data involving
the use of emoticons, emojis, and stickers, in order
to be able to map the evolutionary trajectory of the
three types of graphicons.
2.2 Graphicon on Chinese social media
Graphicons on Chinese social media are
distinctive in their design and usage. They are
designed in creative ways by and for Chinese
social media users to enliven conversations (Ge,
2020), resolve the tension between the openness
of social media and constraint-bounded social
norms (Zhang et al., 2021), and playfully subvert
reality and avoid internet surveillance and
censorship (Li & Zhu, 2019). The design of
graphicons carries rich cultural messages (de Seta,
2018) and interacts with the Chinese national
character (Li & Zhu, 2019).
Users of Chinese social media use the umbrella
term 表情 Biaoqing (a contraction of 表达情感
‘expressing emotions’) for all types of graphicons,
suggesting a popular understanding of the shared
usage of graphicons for emotion expression (de
Seta, 2018). Yet different types of Biaoqing are
distinguished. Kaomojis were introduced to
Chinese users in the mid-1990s; emojis were first
used in the early 2000s in Chinese discussion
boards, instant messaging services, and social
networking web sites; and stickers first became
available on the QQ and WeChat platforms in
2012 (de Seta, 2018). As in the West, all three
types of graphicons are currently available for use.
In terms of frequency of use, Konrad et al.
(2020) suggest that graphicon evolution is more
advanced in Asia than in the West. They predict
that stickers should be catching up with or
surpassing emoji use in Asia, in contrast to the
West, where stickers are still much less popular
than emojis. The evidence to support this
prediction so far is limited and primarily
anecdotal. Fifteen years ago, Markman and
Oshima (2007) reported that the use of kaomojis
as punctuation was more conspicuous in Japan
than the United States. Emojis are used more
frequently by Chinese social media users
compared to their Western counterparts (Zhang et
al., 2014); however, comparable statistics about
the frequency of sticker use have not been found.
Several studies have pointed out that stickers are
now very popular among Chinese social media
users (e.g., de Seta, 2018; Ge, 2020), but their
frequency has not been compared with that of
emojis. In this paper, we quantify the relative
frequency of the three different types of
graphicons in Chinese social media over time.
2.3 Research questions
Based on the gaps delineated in the above
literature review, this study addresses the
following research questions:
RQ1: What are the relative frequencies of each
type of Chinese graphicon, and how have
their frequencies changed over time?
RQ2: What trends are evident from the most
frequently-used graphicons of each type?
3 Methodology
3.1 Data
Our corpus is composed of 13 years of
longitudinal data from the BiliBili platform.
BiliBili is a video-sharing platform that, like
YouTube, allows users to post comments below
the videos and also features short danmu
messages that are overlaid on the video itself.
The BiliBili platform was chosen for several
reasons. First, it is one of the most popular
Chinese social media platforms. The users of the
platform are mainly under the age of 35,
2
and its
average monthly active users reached 272 million
(almost one-fifth of the Chinese population) by
the end of 2021.
3
Second, the comments can
include emoticons, emojis, and stickers (although
stickers did not become available on the platform
until 2016). Third, BiliBili is well-established,
having been launched in 2009 as a platform for
sharing ACG-related (Anime, Comics, and Games)
content,
4
and it has expanded over the years to
cover more general topics. Last and most relevant
for this study, the platform preserves a historical
record of the comments posted below the videos,
including the graphicons in the comments, and the
comments can be captured automatically. We
considered other popular Chinese social media
platforms (e.g., Sina Weibo, WeChat) as possible
data sources, but none of them would have
allowed automatic capturing of longitudinal data
containing all three graphicon types.
The data consist of comments and replies to
comments (hereafter, replies) from the channel of
BiliBili’s annual Spring Festival Gala Show
2
https://socialbeta.com/t/reports-bilibili-marketing-
planning-2021-02-22, retrieved April 4, 2022.
3
https://m.jiemian.com/article/7167482.html, retrieved
April 4, 2022.
4
https://zh.wikipedia.org/wiki/Bilibili, retrieved April 4,
2022.
(hereafter, the BiliBili show).
5
This channel was
chosen because it is the only one that includes
comments dating back to 2010, and the comments
are all on videos on the same topic. The BiliBili
show started in 2010 and soon became an
important annual event on the platform.
6
The
show consists of a mash-up video of content
provided by professional users to celebrate the
Chinese New Year, and is released on the eve of
the Chinese New Year. It is considered by BiliBili
users to be the online equivalent of the Spring
Festival Gala produced by the China Media Group,
which is broadcast annually on Chinese New
Year’s Eve and has the largest audience of any
entertainment show in the world. Besides the
show videos, the BiliBili show channel includes a
number of videos related to the gala show, such as
trailers, teasers, and outtakes. These videos
include older comments and replies, like the gala
show videos do, and are thus included in our data.
Comments and replies from all 42 videos
available in the channel, covering the years from
2010 to 2022, were captured and stored in
February 2022 using Python and the Scrapy tool.
A total of 941,020 messages (including both
comments and replies) were collected.
3.2 Methods
The three types of graphicons in the corpus were
identified using different methods. The emoticons
in our corpus are Japanese-style kaomoji. The
recognition of kaomojis was carried out by a semi-
supervised process of deep learning and manual
identification. Manual annotation of kaomojis in a
sample corpus was done, and this was used to train
deep learning models of BiLSTM and CRF (Qin
et al., 2019) to learn and develop a list of kaomoji
types. Kaomoji types identified by the algorithm
were checked manually. Three rounds of manual
and machine iteration were conducted before a
final set of kaomoji types was obtained for the
purpose of examining kaomoji use in the corpus.
The set of sticker types was developed based
on the package of BiliBili stickers available on
GitHub.
7
The set of Yellow Faces [小黄脸 ] from
the GitHub sticker package (see Figure 1)
5
https://space.bilibili.com/1868902080
6
https://www.bilibili.com/read/cv1069082, retrieved April
5, 2022.
7
https://github.com/amtoaer/bilibili-stickers, retrieved
February 25, 2022. The GitHub collection was updated on
January 31, 2022; all comments and replies in our data were
Figure 1: Examples from the set of Yellow Faces.
contains a number of graphicons that we
reclassified as emojis, as described below. The set
includes three kinds of images: 1) iconic
representations of objects (e.g., Koi [锦鲤], the
second from the top left in Figure 1); 2) yellow
faces that are more elaborated than Unicode
emojis (e.g., Astonished face [惊讶], the second
from the top right; and 3) stickers that are
character-driven (e.g., the Laigu [来古] series of
three girls expressing contemplation [沉思] (the
third from the bottom right), dullness [ 呆滞]
(second from the bottom right), and doubt [疑问]
(bottom right). However, the iconic images and
yellow faces are displayed in the corpus the same
size as emojis, which are smaller than stickers,
and they are not character-driven. Therefore, we
removed them from the sticker package and added
them to the set of emojis prior to analysis. The set
of emojis also includes Unicode emojis from the
Python emoji module with a character length of 1.
Using the above methods, a list of the three
types of graphicons was derived for obtaining
graphicon occurrences in the corpus. The
frequencies of graphicon types and tokens in each
year were obtained. We also conducted a thematic
content analysis of the 10 most frequently
occurring graphicons of each type in the corpus.
4 Findings
The findings are presented in two parts. The first
part reports the frequency distribution of the
graphicons over time. The second part presents a
qualitative analysis of the most frequently used
graphicons in terms of what they suggest about
trends in Chinese graphicon evolution.
4.1 Frequency distribution of graphicons
Three types of graphicon were identified in the
corpus: kaomoji, emoji and sticker. Definitions
and descriptions of each types are provided in the
made after that. We manually confirmed that the Github
package included all the stickers in our corpus.
Figure 2: Frequencies (tokens) of three graphicons.
Figure 3: Frequencies (types) of three graphicons.
Figure 4: Frequencies of messages containing at least
one graphicon.
Introduction and in Section 3.2.
The frequency of each graphicon type was
normalized as a ratio in relation to the number of
messages in the corpus. This was done because
some messages lack text and include only
graphicons. Normalized frequencies of all
graphicon tokens for each of the 13 years are
shown in Figure 2, and normalized frequencies of
graphicon types are in Figure 3.
These statistics provide partial support for the
evolutionary trajectory proposed by Konrad et al.
(2020). The use of kaomojis shows a clear
trajectory of an early phase, a high phase, and
decline. The peak of kaomoji tokens appears in
2016, and the peak of kaomoji types comes earlier
in 2013.
Moreover, kaomojis have been replaced by
emojis and stickers. Emojis experienced a
dramatic increase in occurrences in the most
recent three years (the red bars in 2020, 2021 and
2022 in Figure 2), and the types of emoji also
show a notable uptick in 2022 (the red bars in
2020, 2021 and 2022 in Figure 3). The picture for
stickers is somewhat less clear. After stickers
appeared on BiliBili in 2016, their usage increased
and rose sharply in 2021. Although the frequency
of sticker tokens dropped off in 2022, the types
increased steeply. That is, fewer stickers were
used in 2022 than in 2021, yet many more
varieties of stickers appear in the 2022 data.
It is possible that stickers use has started to
decline in 2022. But it is also possible that the
sharp rise in 2021 is due to unconventional usage
of graphicons by the large number of new users
who joined BiliBili during the Covid-19
pandemic. User numbers increased by 55% to 202
million in 2020,
8
as a result of intensive branding
promotion of the platform.
9
The new users might
have initially used stickers more frequently than
older users but gradually accommodated their
graphicon use to the norms of the community.
The frequencies of messages (comments or replies)
that contain at least one of the three types of
graphicon is shown in Figure 4. The nearly
identical pattern of kaomojis in Figure 2 and
Figure 4 suggests that a kaomoji was mostly used
only once per message. However, two or more
emojis are commonly used in a message. In the
statistics from 2022, for example, 55 emojis
appear per 100 messages (the red bar in 2022 in
Figure 2), but these emoji only appear in 33% of
the messages (the red bar in 2022 in Figure 4).
Stickers tend to be used once per message in early
years (see the similar frequencies in 2016-2020),
but they are used on average more than once in
2021 (40 stickers appear per 100 messages, but
they appear in only 22% of the messages).
Further, graphicon usage differs in comments
and replies, as summarized in Figure 5. Kaomojis
and stickers appear more frequently in comments,
while emojis are used more frequently in replies.
8
Graphicon usage of commenters who joined in 2020 is
reflected in the data of 2021. This is because the show was
released in January 2021, and a majority of comments was
made within the first month of the video release.
Figure 5: Frequencies (tokens) of three graphicons in
comments and replies.
Note: The calculation excludes the data from the years 2010
and 2011, since there were no replies in those years.
Figure 6: Frequencies of graphicon use in comments
and replies.
Note: No replies were made in 2010 and 2011.
Figure 7: Frequencies of comments and replies
containing at least one kaomoji.
Meanwhile, as displayed in Figure 6, a
consistent pattern is found whereby more
graphicons were used in comments than in replies
every year except for 2022. It is also worth noting
that stickers were available on the platform in
2016, but they were not used in replies until 2018;
the reasons for this lag are unclear.
9
https://www.sohu.com/a/452506920_153054, retrieved
April 7, 2022.
Figure 8: Frequencies of comments and replies
containing at least one emoji.
Another interesting phenomenon is the strong
correlation between the frequency of each
graphicon type in the comments and their
corresponding replies. This is evident for
kaomojis in Figure 7 and for emojis in Figure 8.
Frequencies for stickers are not presented here
because stickers were not available on BiliBili
until 2016. The years of the replies in Figures 7 &
8 refer to the year when the corresponding
comments were made rather than the year when
the replies were made (as in Figure 6). For
instance, for a reply made in 2022 to a comment
from 2010, the year of the reply was counted as
2010 in Figures 7 & 8 but counted as 2022 in
Figure 6. The frequencies of kaomoji usage in
comments and replies (Figure 7) show a very
similar pattern; the correlation is 0.95. The
frequencies of emojis (Figure 8) show a less
consistent pattern, but the correlation between
comments and replies is still strong at 0.88. The
strong correlations between graphicon usage in
comments and replies suggest a “priming effect”
(Molden, 2014) of graphicon usage, meaning that
the occurrences of graphicons in comments have
an impact on the usage of graphicons in
corresponding replies.
4.2 The top 10 graphicons
Next, we qualitatively examined the most
frequently occurring graphicons in the corpus.
The top 10 occurrences of each graphicon in each
category are listed in Table 1. In general, the
progression from kaomojis to emojis to stickers
reveals a trend of movement from general
emotion expression to meanings localized in the
discourse practices of the BiliBili platform.
10
https://www.bilibili.com/read/cv1069082, retrieved April
5, 2022.
Kaomojis are borrowed from Japanese to
express emotions, and indeed, the most frequently
used kaomoji types in our corpus mainly express
emotion. Four kaomojis express joy (Nos. 3, 4, 7
& 10). Five kaomojis perform actions with
incorporated affect (Nos. 1, 5, 6, 8 & 9). There is
no explicit encoding of affect in the kaomojis of
cheering (No. 1) or dancing with music (No. 6),
but these two actions are strongly conventionally
associated with a happy mood.
In contrast, fewer of the 10 most popular emojis
focus on emotions. Rather, several of the emojis
reference culture-specific information about the
New Year’s celebration event and the BiliBili
platform. Three emojis (Nos. 2, 3 & 8) are for the
Chinese new year celebration. The emoji of year
of the rat (No. 3) integrates the shape of a TV set,
the icon that symbolizes BiliBili, in their design
(see Nos. 3, 5 & 9 in stickers). It is worth noting
that only one of the popular emojis are Unicode
emojis (No. 8, Clinking beer mugs), supporting
previous findings that platform-specific sets of
graphicons are more popular in China than
Unicode emojis (de Seta, 2018; Y. Zhang et al.,
2021). Even the Unicode (clinking beer mugs)
emoji is localized in meaning, in that it is
frequently used in Chinese New Year’s wishes as
a symbolic representation of cheering.
Integration with platform discourse practices is
most evident in stickers. Three stickers belong to
the Popular Words Series (Nos. 1, 2 & 7), which
are graphic representations of selected popular
expressions from comments or danmu (messages
that are overlaid on the video itself). In addition,
the stickers include three variants of the
platform’s icon (Nos. 3, 5 & 9 in the “Tiny TV”
set) and two virtual spokespersons of BiliBili
10
(Nos. 6 & 10 in the “2233 Girls” set).
The themes expressed by the three types of
graphicons are summarized in Table 2. Emotion
expression becomes less prominent as we move
from kaomoji to emoji and to sticker. In contrast,
references to platform discourse and the
integration of Chinese characters become more
apparent as we move from the older to the newer
graphicons. Relatedly, action decreases somewhat.
For most of the themes, emoji serves as a
transition between kaomoji and sticker.
Table 1: Top 10 graphicons in the corpus.
Notes: 1) The meanings of most of the kaomojis were derived by referring to the kaomoji dictionary (Kaomoji-Japanese
Emoticons, http://kaomoji.ru/en/). The first kaomoji is not found in the dictionary; its meaning was derived from the fact that
it is always used together with the Chinese expression
干杯
‘cheers/toast’. 2) The kaomojis of Nos. 5, 6, 8 and 9 are a
combination of at least two kaomoji elements from the kaomoji dictionary. For example, [°
∀
°] refers to joy, while [
ノ
] refers to
a hand waving to greet someone, suggesting that the meaning of [(°
∀
°)
ノ
] is greeting happily. 3) PWS in the sticker category
is short for Popular Word Series, the meanings of which can be found on the BiliBili platform:
https://www.bilibili.com/read/cv4332187
Kaomoji
Emoji
Sticker
Emotion
70%
30%
20%
Action
50%
40%
40%
Chinese character
0
10%
30%
New Year’s celebration
10%
30%
20%
Platform discourse
10%
90%
100%
Table 2: Themes of the top 10 graphicons.
Kaomoji
Emoji
Sticker
Graph-
icon
(⌒▽⌒)
Table 3: Graphicon evolution for smile.
The trajectory from generalized emotion
expression to localized platform discourse
practices is illustrated by the example of ‘smile’
in Table 3. The kaomoji represents smile in an
abstract and general way, using ⌒ to indicate
eyes and ▽ for nose. The emoji smile is different
from the smiles on other Chinese social media
platforms such as Weibo and WeChat, but still it
is somewhat generic and does not encode any
platform information. In contrast, the sticker
smile is unique to BiliBili, in that it is embedded
in the BiliBili icon of a tiny TV set.
Another example is the concept ‘wonderful,’
which is expressed with a dog face emoji (No. 4
in emojis; see Table 1) but represented by a
combination of the Chinese character 妙 and an
exclamation point, an example of the Popular
Words Series set of stickers (No. 1 in stickers in
Table 1). These examples illustrate that the
discourse practices of the platform have
increasingly been encoded in graphicons.
5 Discussion
5.1 Research questions revisited
We asked how the frequencies of each of the
three types of Chinese graphicons are changing
over time and what trends are evident from the
most frequently-used graphicons of each type.
The use of kaomojis shows a clear trajectory of
rising to a peak and then declining. Kaomojis
have been replaced by emojis and stickers. These
results support Konrad et al.’s evolutionary
trajectory. It is less clear, however, whether
stickers are overtaking emoji in frequency of use
on BiliBili; rather, both appear to be on the rise.
Moreover, in the last three years (2020-2022),
emojis were used with high frequency but with a
limited number of types, and it is common to find
more than one emoji in a message. As for stickers,
there is a decrease in tokens but an increase in
types. We propose that frequency of type be
included as a criterion for determining which
phase a graphicon is currently in.
The trend revealed by the most frequently-
used graphicons of each type suggests an
evolution from general emotion expression to
meanings localized in platform discourse
practices. This supports Konrad et al.’s (2020)
finding that stickers express more specific
meanings than emojis.
We also find an increasing integration of
Chinese characters in emojis and stickers. The
logographic nature of Chinese characters (Li &
Zhu, 2019) makes such integration possible.
Though we did not find integration of Chinese
characters in the most frequently-used kaomojis,
Chinese users in the 1990s were inspired by the
practice of using keyboard symbols in kaomojis
to create unique graphic representations of
Chinese characters for festival celebrations, as
shown in the examples in Kozar (1995).
These findings suggest a somewhat different
evolutionary trend than that for Western
graphicons proposed by Konrad et al. (2020).
The types and tokens of emojis and stickers are
both on the rise, although stickers do not seem to
be overtaking emojis. It is highly possible that
emojis have not reached their peak yet.
Meanwhile, features of stickers, such as specific
references and more detailed graphics, are
increasingly being incorporated into emoji
design (e.g., the Astonished face in Figure 1). If
this trend continues, it is likely to expand the
functions of emojis and blur the distinction
between emojis and stickers. The icons in the set
of Yellow Faces in the GitHub package of
BiliBili stickers that we reclassified as emojis (as
discussed in Section 3.2) are somewhat
ambiguous between the two graphicon types.
Meanwhile, the fact that more than one sticker is
used per message suggests that users are
borrowing from emojis the practice of repeating
graphicons in one message. Thus the interrelation
of emojis and stickers, as the examples and
statistics in this study show, is more complex
than one replacing the other.
5.2 Unanticipated findings
Unexpectedly, our results showed different
patterns of graphicon usage in comments and
replies. More graphicons were used in comments
than in replies overall. This finding differs from
that of Kaneyasu (2022), who conducted a
qualitative study of the use of kaomojis in a
Japanese user-generated recipe sharing site.
Kaomojis appeared more frequently in replies
that were directed at individuals than in
comments directed at more general readers. It
remains to be explored further using both
qualitative and quantitative methods how and
why graphicon are used in different ways in
comments and replies.
Furthermore, we found that more kaomojis
and stickers were used in comments, but the use
of emojis was roughly the same in comments and
replies. At the moment, we do not have a
plausible explanation for this finding, but it at
least suggests that certain properties are shared
between kaomojis and stickers. This
phenomenon also requires further study.
Last, our statistics suggest a “priming effect”
of graphicon usage in comments and replies. The
use of kaomojis and emojis in replies shows
strong correlations with the occurrences of these
two types of graphicon in their corresponding
comments. Emotion expressions tend to
demonstrate priming effects (e.g., Neumann,
2000), but studies about the priming effects of
graphicons have focused mainly on the functions
of graphicons as primes on language use and
processing. For instance, it has been found that
emoji primes function as paralanguage to
facilitate the processing of relevant emotive
linguistic expressions (Yang et al., 2021). Our
findings provide evidence that priming is taking
place as regards graphicon forms.
6 Conclusion
6.1 Contributions
This paper makes several novel contributions. It
presents what we believe is the first longitudinal,
comparative study of graphicon use on a Chinese
social media platform. It provides support for
Konrad et al.’s (2020) evolutionary model
concerning the relationship between emoticons
(kaomoji) and other graphicons. However, the
BiliBili data do not show stickers leading or
taking over from emoji, contrary to Konrad et
al.’s intriguing speculation that Chinese
graphicons would show that trend. Moreover, our
qualitative analysis of the top 10 most frequently
used graphicons reveals a trend of graphicon
evolution from general emotion expression to
meanings localized in the discourse practices of
the BiliBili platform.
Further, our analysis of a large longitudinal
dataset went beyond reporting overall
frequencies of occurrence to explore more fine-
grained distinctions between types and tokens
and differences in graphicon usage between
comments and replies to comments. We also
provided statistical evidence for priming effects
on graphicon usage in comments and replies.
These contributions reveal the complexities of
graphicon evolution on BiliBili and generate
additional research questions.
6.2 Limitations
A number of limitations potentially affect the
generalizability of the patterns of graphicon
evolution identified in this study. First, trends in
Chinese graphicon usage might differ on a
different platform such as WeChat or Weibo,
because Chinese social media users are inclined
to use platform-specific sets of graphicons.
Kaomoji usage is likely more frequent on BiliBili
than on any other platform, given that the
platform was initially set up to share Japanese
anime, comics, and games. Stickers from users’
collections cannot be used on BiliBili like they
are on WeChat and Weibo. Meanwhile, only a
very limited number of sticker sets are free to
users, which is likely to have an impact on the
varieties of stickers in use. For instance, all of our
top 10 stickers are from free sets rather than paid
ones. The landscape of sticker usage in WeChat
or Weibo, therefore, could be very different.
Meanwhile, a majority of BiliBili users are under
the age of 35, and this demographic might use
graphicons differently from older groups.
Second, our data center on the topic of the
Chinese New Year, which is both a strength and
a limitation of our study. The topic provided
straightforward clues for interpreting the
meaning of graphicons, and the fixed content
allowed us to focus on graphicon forms.
However, while kaomoji meanings are rather
general, the denotations of emojis and stickers
are increasingly content specific; thus their usage
might vary for different topics. We would not
expect, for instance, to find as many graphicons
on the theme of the Chinese New Year in
comments on videos on other topics. More topics,
and different platforms, should therefore be
analyzed in order to increase the generalizability
of our evolutionary findings.
Another factor that might have impacted the
evolutionary trajectory is the limited data from
2022. In order to have as much longitudinal data
as possible, we included data from 2022;
however, these were from only the first two
months of the year, so the number of messages
from 2022 is relatively small compared with the
preceding years. We therefore should be cautious
in interpreting the statistics about graphicon
usage in 2022, as they might not fully represent
the graphicon usage of the year. Follow-up study
with future data from BiliBili is needed to
develop a fuller picture of graphicon evolution on
the platform, particularly with regard to emojis
and stickers.
6.3 Future directions
The findings from this study suggest a number of
directions for further research. First, a more
detailed description of graphicon evolution could
be obtained by establishing a relationship
between graphicon usage and user demographics
such as gender. Second, the differences in
graphicon usage between comments and replies
could be investigated further by examining the
pragmatic functions of the graphicons and their
positions in sentences. Qualitative analysis could
also shed light on how and why the priming
effect takes place in graphicon usage.
Last, as Chinese language features are
increasingly integrated with graphicons, it is
important to examine the impact of graphicons
on textual language and language use.
Pavalanathan and Eisenstein (2016) found that
creative spelling and typography decreased on
Twitter as emoji use increased. We have
informally observed a decrease in the use of
Chinese words that express attitude on BiliBili as
graphicon use has increased over time. This
suggests that as graphicons evolve, they are not
just supplementing text but are partially
replacing it. A study of graphicon frequencies in
relation to word frequencies at different points in
time could provide empirical evidence in support
of this proposition.
Acknowledgments
The authors would like to thank Xixuan Huang,
Rongle Tan, and Yanmin Wu for their help with
data cleaning, annotation, and analysis. This
study was funded by Guangdong Planning Office
of Philosophy and Social Science (Grant No.
GD21CWY07), and the Department of
Education of Guangdong Province (Grant No.
2020WTSCX016).
References
Fathiya Al Rashdi. 2018. Functions of emojis in
WhatsApp interaction among Omanis.
Discourse, Context & Media, 26: 117-126.
https://doi.org/10.1016/j.dcm.2018.07.001
Marcel Danesi. 2016. The Semiotics of Emoji: The
Rise of Visual Language in the Age of the
Internet. Bloomsbury Academic, London &
New York.
Gabriele de Seta. 2018. Biaoqing: The circulation of
emoticons, emoji, stickers, and custom images
on Chinese digital media platforms. First
Monday, 23(9). 10.5210/fm.v23i9.9391
Vyvyan Evans. 2017. The Emoji Code: The
Linguistics behind Smiley Faces and Scaredy
Cats. Picador, New York.
Jing Ge. 2020. The anatomy of memetic stickers: An
analysis of sticker competition on Chinese social
media. In Proceedings of the ICWSM
Workshops, pages 1-8.
https://doi.org/10.36190/2020.01
Jing Ge and Susan C. Herring. 2018. Communicative
functions of emoji sequences on Sina Weibo.
First Monday, 23(11).
http://dx.doi.org/10.5210/fm.v23i11.9413
Susan C. Herring and Ashley Dainas. 2017. "Nice
picture comment!" Graphicons in Facebook
comment threads. In Proceedings of The Fiftieth
Hawai'i International Conference on System
Sciences. IEEE Press, pages 2185-2194.
https://doi.org/10.24251/HICSS.2017.264
Michiko Kaneyasu. 2022. Multimodal strategies for
balancing formality and informality: The role of
kaomoji in online comment-reply interactions.
Internet Pragmatics, 5(1): 143-164.
https://doi.org/10.1075/ip.00071.kan
Hirofumi Katsuno and Christine R. Yano. 2002. Face
to face: on‐line subjectivity in contemporary
Japan. Asian Studies Review, 26(2): 205-231.
https://doi.org/10.1111/1467-8403.00025
Artie Konrad, Susan C. Herring, and David Choi.
2020. Sticker and emoji use in Facebook
messenger: Implications for graphicon change.
Journal of Computer-Mediated Communication,
25(3): 217-235.
https://doi.org/10.1093/jcmc/zmaa003
Seana Kozar. 1995. Enduring traditions, ethereal
transmissions: Recreating Chinese New Year
celebrations on the Internet. Journal of
Computer‐Mediated Communication, 1(2).
https://doi.org/10.1111/j.1083-
6101.1995.tb00329.x
Wei Li and Hua Zhu. 2019. Tranßcripting: playful
subversion with Chinese characters.
International Journal of Multilingualism, 16(2):
145-161.
https://doi.org/10.1080/14790718.2019.1575834
Lorenzo Logi and Michele Zappavigna. 2021. A
social semiotic perspective on emoji: How emoji
and language interact to make meaning in digital
messages. New Media & Society.
https://doi.org/10.1177/14614448211032965
Kris M. Markman and Sae Oshima. 2007. Pragmatic
play? Some possible functions of English
emoticons and Japanese kaomoji in computer-
mediated discourse. Paper presented at the
Association of Internet Researchers Annual
Conference 8.0. Vancouver, Canada.
Daniel C. Molden. 2014. Understanding priming
effects in social psychology: An overview and
integration. Social Cognition, 32(Supplement):
243-249.
https://doi.org/10.1521/soco.2014.32.supp.243
Roland Neumann. 2000. The causal influences of
attributions on emotions: A procedural priming
approach. Psychological Science, 11(3): 179-
182. https://doi.org/10.1111/1467-9280.00238
Umashanthi Pavalanathan and Jacob Eisenstein.
2016. More emojis, less :) The competition for
paralinguistic function in microblog writing.
First Monday, 22(11).
https://doi.org/10.5210/fm.v21i11.6879
Robert R. Provine, Robert J. Spencer, and Darcy L.
Mandell. 2007. Emotional expression online:
Emoticons punctuate website text messages.
Journal of Language and Social Psychology,
26(3): 299-307.
https://doi.org/10.1177/0261927X06303481
Ya Qin, Guowei Shen, Wenbo Zhao, Yanping Chen,
Miao Yu, and Xin Jin. 2019. A network security
entity recognition method based on feature
template and CNN-BiLSTM-CRF. Frontiers of
Information Technology & Electronic
Engineering, 20(6): 872-884.
https://doi.org/10.1631/FITEE.1800520
Agnese Sampietro. 2019. Emoji and rapport
management in Spanish WhatsApp chats.
Journal of Pragmatics, 143: 109-120.
https://doi.org/10.1016/j.pragma.2019.02.009
Ying Tang and Khe Foon Hew. 2018. Emoticon,
emoji, and sticker Use in Computer-Mediated
Communications: Understanding its
communicative function, impact, user behavior,
and motive. In L. Deng, W. W. K. Ma, & C. W.
R. Fong (Eds.), New Media for Educational
Change. Springer, Cham. pages 191-201.
Jiayi Yang, Yingxi Yang, Lichao Xiu, and Guoming
Yu. 2021. Effect of emoji prime on the
understanding of emotional words–evidence
from ERPs. Behavior & Information
Technology: 1-10.
https://doi.org/10.1080/0144929X.2021.1874050
Xiaochen Zhang, Weiting Tao, and Sora Kim. 2014.
A comparative study on global brands’ micro
blogs between China and USA: Focusing on
communication styles and branding strategies.
International Journal of Strategic
Communication, 8(4): 231-249.
https://doi.org/10.1080/1553118X.2014.886251
Yiqiong Zhang, Min Wang, and Ying Li. 2021. More
than playfulness: Emojis in the comments of a
WeChat official account. Internet Pragmatics,
4(2): 247-271.
https://doi.org/10.1075/ip.00048.zha