BilingualChildEmo — XLM-RoBERTa for Bilingual Emotion Detection in Children's Literature

Model Details

Model Description

This model is a fine-tuned version of XLM-RoBERTa for bilingual (Chinese–English) sentence-level emotion classification in children's literature. It classifies text into five basic emotion categories adapted from Parrott (2001): JOY, SADNESS, ANGER, FEAR, and LOVE.

The model was developed as part of the study "Examining emotions in English and translated Chinese children's literature: a bilingual emotion detection model based on LLMs", published in Language Resources and Evaluation (Springer, 2025).

• Developed by: Yanjin Liu, Sophia Yat Mei Lee, Dechao Li
• Funded by: Hong Kong Research Grants Council (RGC) General Research Fund (GRF), Project No. 15611021
• Shared by: Department of Chinese and Bilingual Studies, The Hong Kong Polytechnic University; Faculty of Humanities and Social Sciences, City University of Macau
• Model type: Text Classification (Transformer-based, fine-tuned XLM-RoBERTa)
• Language(s): Chinese (zh), English (en)
• License: Apache-2.0
• Fine-tuned from: xlm-roberta-base

Model Sources

• Repository: nanaaaa/BilingualChildEmo — DOI: https://doi.org/10.57967/hf/1912
• Paper: Liu, Y., Lee, S. Y. M., & Li, D. (2025). Examining emotions in English and translated Chinese children's literature: a bilingual emotion detection model based on LLMs. Language Resources and Evaluation. https://doi.org/10.1007/s10579-025-09846-z
• Dataset: nanaaaa/BilingualChildEmo Dataset — DOI: https://doi.org/10.57967/hf/4379

---

Uses

Direct Use

This model is designed for sentence-level bilingual emotion classification in Chinese and English literary texts, particularly children's literature. Given a sentence in either language, the model predicts one of five emotion labels:

Label	Description
JOY	Happiness, delight, celebration
SADNESS	Sorrow, grief, disappointment
ANGER	Annoyance, hostility, frustration with blame
FEAR	Anxiety about a real or perceived threat or danger
LOVE	Deep affection, attachment, or care for someone/something

Downstream Use

The model can be applied to:

• Cross-linguistic emotion analysis in translated literary texts
• Computational translation studies examining emotional shifts between source and target texts
• Educational NLP tools analyzing affective content in reading materials for children
• Cross-cultural studies of emotional expression differences between Chinese and English

---

How to Get Started with the Model

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="nanaaaa/BilingualChildEmo"
)

# English example
result = classifier("So overjoyed were they at their deliverance that they laughed aloud.")
print(result)  # Expected label: JOY

# Chinese example
result = classifier("可怜的人儿, 失去了他们唯一的儿子!")
print(result)  # Expected label: SADNESS