Emojinize: Enriching Any Text with Emoji Translations

TL;DR

The paper introduces Emojinize, an automatic system that translates arbitrary English text into emoji sequences using a large language model with context-aware disambiguation and compositional emoji expressions. It evaluates emoji translations via cloze-test–based human and automatic studies, showing that Emojinize substantially improves word-guessability and often outperforms human translations. The authors further explore multi-shot translations, batch translations, multi-word expressions, and cross-language capabilities, revealing both benefits and trade-offs. The work highlights the potential of emoji-based representations for literacy, language learning, and inclusive text understanding, while outlining future directions for feedback-driven refinement and synthetic data generation to mature the emoji language ecosystem.

Abstract

Emoji have become ubiquitous in written communication, on the Web and beyond. They can emphasize or clarify emotions, add details to conversations, or simply serve decorative purposes. This casual use, however, barely scratches the surface of the expressive power of emoji. To further unleash this power, we present Emojinize, a method for translating arbitrary text phrases into sequences of one or more emoji without requiring human input. By leveraging the power of large language models, Emojinize can choose appropriate emoji by disambiguating based on context (eg, cricket-bat vs bat) and can express complex concepts compositionally by combining multiple emoji (eq, "Emojinize" is translated to input-latin-letters right-arrow grinning-face). In a cloze test--based user study, we show that Emojinize's emoji translations increase the human guessability of masked words by 55%, whereas human-picked emoji translations do so by only 29%. These results suggest that emoji provide a sufficiently rich vocabulary to accurately translate a wide variety of words. Moreover, annotating words and phrases with Emojinize's emoji translations opens the door to numerous downstream applications, including children learning how to read, adults learning foreign languages, and text understanding for people with learning disabilities.

Figures (7)

• Figure 1: LLM-based translation of words to emoji using few-shot prompting with JSON-formatted assistant output.
• Figure 2: Evaluation protocol. Text from a custom corpus is translated to emoji language. Translations are used in a cloze test. Different translation mechanisms lead to different test conditions. For the analysis we rely on LLM-based scoring.
• Figure 3: A custom emoji picker for allowing humans to translate target words (here: "bed") to emoji (Sec. \ref{['sec:human_eval']}).
• Figure 4: UI of cloze test for human evaluation.
• Figure 5: Accuracy of guessing the hidden word: (a) human participants (Sec. \ref{['sec:Human results']}), (b) LLM participants (Sec. \ref{['sec:Automatic results']}).