Tower of Babel in Cross-Cultural Communication: A Case Study of \#Give Me a Chinese Name\# Dialogues During the ``TikTok Refugees'' Event

Paper Title

Tower of Babel in Cross-Cultural Communication: A Case Study of #Give Me a Chinese Name# Dialogues During the 'TikTok Refugees' Event

Publication Info

• Topic area: Cross-cultural communication and naming practices in digital environments.
• Keywords: Cross-cultural communication, naming practices, social media, human-in-the-loop, large language models, RedNote, TikTok refugees, semantic strategies, phonetic strategies, visual strategies.

Background and Problem

• Problem / challenge: Existing research on social media communication largely focuses on monolingual or single-cultural contexts, leaving cross-cultural and multilingual dynamics underexplored. Naming practices, as a dense cultural and social act, remain particularly challenging for machine translation due to layered semantic, phonetic, and cultural dimensions.
• Significance: Understanding cross-cultural naming practices is critical for improving digital communication, fostering inclusivity, and designing platforms that support multicultural interactions.
• Motivation and related work: Prior studies in HCI have examined user behavior on platforms like Instagram, TikTok, and Weibo, but these are often limited to single-language contexts. Cross-cultural naming practices, such as those observed during the "TikTok refugee" event, provide a unique opportunity to explore how cultural boundaries are negotiated in digital spaces.

Solution

• Proposed approach: A human-in-the-loop method combining large language models (LLMs) and human judgment to analyze cross-cultural naming practices on RedNote, focusing on semantic, phonetic, and visual channels.
• Novelty:
  1. Development of a scalable human–AI collaborative framework for extracting and analyzing cross-cultural naming strategies.
  2. Creation of a systematic three-channel framework (semantic, phonetic, visual) for naming practices.
  3. Insights into how naming strategies combine to construct a "Babel Tower" effect in cross-cultural communication.
  4. Design and methodological implications for multicultural platform governance and AI-based translation tools.
• Procedure and key techniques:
  • Data collection: 318 posts and 70,614 comments from RedNote were analyzed.
  • Human-in-the-loop extraction: Iterative refinement of LLM outputs using consistency-guided human correction.
  • Explanation generation: Semantic, phonetic, and visual explanations were generated for names using structured prompts and LLMs.
  • Framework development: A three-channel framework of naming strategies was synthesized through clustering and manual validation.
  • Engagement analysis: Likes were analyzed to understand user preferences and visibility dynamics.

Results

• Concrete findings:
  • 31 semantic naming strategies were identified, organized into four overarching categories: Foreign Cultural References (15.78%), Chinese Traditional Culture (40.66%), Chinese Social Identity (13.64%), and Chinese Internet Pop (29.93%).
  • Phonetic strategies (30.4%) included full homophony (40.54%), partial homophony (50.67%), and meaning association (8.79%).
  • Visual strategies (20.67%) linked names to attributes like demeanor (31.13%), facial features (21.52%), and hair (17.84%).
  • Multi-channel combinations were rare (3.57%), with single-channel strategies dominating (64.07%).
  • Engagement patterns showed that National Narratives, Miscellaneous Memes, and Celebrities attracted the most likes.
• Advantage over baselines: The human-in-the-loop approach achieved 99.66% accuracy in name-explanation extraction, significantly outperforming purely manual or automated methods.
• Experiments / evaluation:
  • Data: 318 posts and 70,614 comments from RedNote.
  • Metrics: Accuracy of name-explanation extraction, engagement (likes), and clustering validation (Cohen’s κ = 0.77 for semantic categories).
  • Methods: Iterative refinement of LLM prompts, clustering (BERTopic), and manual validation by experts.
• Limitations and future work:
  • Dataset may be biased toward highly visible posts due to sampling methods.
  • Limited access to nested replies and deeper conversational structures.
  • Future research should explore other platforms, cultural contexts, and symbolic practices like memes and multimodal discourse.

Summary

This paper investigates cross-cultural naming practices on RedNote during the "TikTok refugee" event, where Chinese users assigned names to foreign newcomers. A human-in-the-loop approach leveraging LLMs and manual correction was developed to analyze 70,614 comments, resulting in a three-channel framework of naming strategies (semantic, phonetic, visual). The study reveals how these strategies combine to create a "Babel Tower" effect, complicating cross-cultural communication. Engagement analysis highlights the prominence of humor, parody, and cultural references in fostering interaction. These findings have implications for platform design, AI-based translation tools, and the governance of multicultural digital spaces.