TL;DR
This study analyzes 13 years of Bilibili comments to track the evolution of emoticons (kaomoji), emojis, and stickers. It reveals a shift from general emotion expression to platform-specific cultural meanings, with kaomoji declining, while emojis and stickers rise, suggesting a complex interplay rather than simple replacement.
What Problem Does This Paper Address?
The way we communicate online is constantly evolving, and graphical icons, or "graphicons," play a huge role in this. From the early text-based emoticons like :-) to the rich visual language of emojis and stickers we use today, these icons have changed dramatically. While previous research has explored the individual use and meaning of these graphicon types, there's a significant gap in understanding their evolutionary trajectory, especially in non-Western contexts. Specifically, how do these different types of graphicons compete, complement, or replace each other over time? This paper tackles this by examining the longitudinal usage of kaomoji, emojis, and stickers on Bilibili, a major Chinese video-sharing platform, to understand their dynamic evolution and the cultural nuances shaping their use.
Methods
The researchers employed a multi-faceted approach to analyze graphicon usage on Bilibili:
- Data Collection: A 13-year corpus of comments and replies from Bilibili's annual Spring Festival Gala Show channel (2010-2022) was collected using Python and the Scrapy tool, resulting in over 1 million messages.
- Graphicon Identification:
- Kaomoji: Identified through a semi-supervised deep learning process combined with manual annotation to train models for recognizing kaomoji types.
- Emojis: Unicode emojis were identified, along with specific "Yellow Faces" from a Bilibili sticker package that were reclassified as emojis due to their size and schematic nature.
- Stickers: Identified based on a GitHub package of Bilibili stickers.
- Quantitative Analysis:
- Frequency Distribution: The frequency of each graphicon type (tokens and types) was calculated per year, normalized by the total number of messages.
- Comment vs. Reply Analysis: Graphicon usage was compared between comments and replies to identify differences and potential priming effects.
- Qualitative Analysis: A thematic content analysis was performed on the top 10 most frequently occurring graphicons of each type to understand their evolving meanings and cultural integration.
Datasets
The primary dataset used in this study is a corpus of over 1 million messages (comments and replies) collected from Bilibili's annual Spring Festival Gala Show channel, spanning from 2010 to 2022. This dataset is unique in its longitudinal nature and its focus on a specific, high-traffic channel, allowing for the tracking of graphicon evolution over a significant period. The study also references publicly available Bilibili sticker packages from GitHub for sticker identification.
Key Insights and Ideas
- Evolutionary Trajectory Supported, But Nuanced: The study largely supports Konrad et al.'s (2020) evolutionary model of graphicons (early, peak, decline/conventionalization). Kaomoji clearly shows this pattern, declining after a peak. However, the relationship between emojis and stickers is more complex than a simple replacement, with both showing signs of growth.
- Beyond Emotion: Cultural Localization: A significant trend observed is the shift from general emotion expression (especially in kaomoji) towards meanings deeply embedded in Bilibili's platform-specific culture and the Chinese language. Stickers, in particular, are rich with references to popular online expressions, Bilibili's own branding, and cultural events.
- "Priming Effect" in Action: A strong correlation was found between graphicon usage in comments and their corresponding replies. This suggests a "priming effect," where the graphicons used in a comment influence the graphicons chosen in the reply, indicating a form of visual conversation flow.
- Stickers: More Than Just Pictures: Stickers are not just decorative; they are increasingly incorporating text and are often character-driven, reflecting platform-specific discourse and even acting as a form of "paralanguage" that complements or replaces textual elements.
- Ambiguity and Blurring Lines: The reclassification of some "Yellow Faces" from sticker packages as emojis highlights the blurring lines between these categories, especially as emojis begin to adopt more detailed graphics and specific cultural references, mirroring sticker design trends.
Key Results
- Kaomoji Decline: Kaomoji usage (both tokens and types) peaked around 2013-2016 and has been declining since, being largely replaced by emojis and stickers.
- Emoji and Sticker Rise: Emojis have seen a dramatic increase in usage in recent years (2020-2022), with a notable uptick in the variety of emoji types used in 2022. Stickers also showed a sharp increase in usage in 2021, though token frequency dropped in 2022, while types increased.
- Comment vs. Reply Differences: Kaomojis and stickers were found to be more frequent in comments than replies. Emojis, however, showed similar frequency in both.
- Priming Effect Strength: Strong correlations (0.93 for kaomoji, 0.89 for emoji) were found between graphicon usage in comments and replies, providing statistical evidence for a priming effect.
- Cultural Integration: The top 10 graphicons reveal a clear progression: kaomojis primarily express emotion (70% emotion, 50% action), emojis show a mix (30% emotion, 40% action, 40% New Year's celebration), and stickers are heavily integrated with platform discourse and Chinese characters (20% emotion, 40% action, 40% Chinese character, 40% New Year's celebration, 100% platform discourse).
Limitations and Future Work
- Platform Specificity: The study's findings are specific to Bilibili. Trends in graphicon usage might differ on other Chinese social media platforms like WeChat or Weibo, which have different user bases and sticker policies.
- Demographic Influence: The majority of Bilibili users are under 35, and their graphicon usage might differ from older demographics.
- Topic Specificity: The data was collected from a specific channel (Spring Festival Gala Show). While this provided clear context for interpretation, graphicon usage might vary across different video topics.
- Limited 2022 Data: The 2022 data only covers the first two months, making it difficult to draw firm conclusions for the entire year.
- Future Directions: The authors suggest further research into graphicon evolution across different platforms and user demographics, a deeper examination of comment-reply differences and the priming effect, and an investigation into how graphicons are increasingly replacing textual elements in communication.
Summary updated