AnasAlokla/multilingual_go_emotions

🌍 Multilingual GoEmotions Classifier 💬

Table of Contents

• 📖 Overview
• ✨ Key Features
• 💫 Supported Emotions
• 🔗 Links
• ⚙️ Installation
• 🚀 Quickstart: Emotion Detection
• 📊 Evaluation
• 💡 Use Cases
• 📚 Trained On
• 🔧 Fine-Tuning Guide
• 🏷️ Tags
• 💬 Support & Contact

Overview

This repository contains a powerful multilingual, multi-label emotion classification model. It is fine-tuned from the robust bert-base-multilingual-cased model on the comprehensive multilingual_go_emotions dataset. The model is designed to analyze text and identify a wide spectrum of 27 different emotions, plus a neutral category. Its ability to detect multiple emotions simultaneously makes it highly effective for understanding nuanced text from diverse sources.

• Model Name: AnasAlokla/multilingual_go_emotions
• Architecture: BERT (bert-base-multilingual-cased)
• Task: Multi-Label Text Classification
• Languages: Arabic, English, French, Spanish, Dutch, Turkish

Key Features

• 🌍 Truly Multilingual: Natively supports 6 major languages, making it ideal for global applications.
• 🏷️ Multi-Label Classification: Capable of detecting multiple emotions in a single piece of text, capturing complex emotional expressions.
• 💪 High Performance: Built on bert-base-multilingual-cased, delivering strong results across all supported languages and emotions. See the detailed evaluation metrics.
• 🔗 Open & Accessible: Comes with a live demo, the full dataset, and the complete training code for full transparency and reproducibility.
• V1.1 Improved Version: An updated model is available that specifically improves performance on low-frequency emotion samples.

Supported Emotions

The model is trained to classify text into 27 distinct emotion categories as well as a neutral class:

Emotion	Emoji	Emotion	Emoji
Admiration	🤩	Love	❤️
Amusement	😄	Nervousness	😰
Anger	😠	Optimism	✨
Annoyance	🙄	Pride	👑
Approval	👍	Realization	💡
Caring	🤗	Relief	😌
Confusion	😕	Remorse	😔
Curiosity	🤔	Sadness	😢
Desire	🔥	Surprise	😲
Disappointment	😞	Disapproval	👎
Disgust	🤢	Gratitude	🙏
Embarrassment	😳	Grief	😭
Excitement	🎉	Joy	😊
Fear	😱	Neutral	😐

Links

• Live Demo: Hugging Face Space
• Dataset (Supports 6 Languages): multilingual_go_emotions
• Model Used: AnasAlokla/multilingual_go_emotions
• GitHub Code: emotion_chatbot
• Improved Version (V1.1): multilingual_go_emotions_V1.1

Installation

Install the required libraries using pip:

pip install transformers torch

Quickstart: Emotion Detection

You can easily use this model for multi-label emotion classification with the transformers pipeline. Set top_k=None to see all predicted emotions above the model's default threshold.

from transformers import pipeline

# Load the multilingual, multi-label emotion classification pipeline
emotion_classifier = pipeline(
    "text-classification",
    model="AnasAlokla/multilingual_go_emotions",
    top_k=None # To return all scores for each label
)

# --- Example 1: English ---
text_en = "I'm so happy for you, but I'm also a little bit sad to see you go."
results_en = emotion_classifier(text_en)
print(f"Text (EN): {text_en}")
print(f"Predictions: {results_en}\n")

# --- Example 2: Spanish ---
text_es = "¡Qué sorpresa! No me lo esperaba para nada."
results_es = emotion_classifier(text_es)
print(f"Text (ES): {text_es}")
print(f"Predictions: {results_es}\n")

# --- Example 3: Arabic ---
text_ar = "أشعر بخيبة أمل وغضب بسبب ما حدث"
results_ar = emotion_classifier(text_ar)
print(f"Text (AR): {text_ar}")
print(f"Predictions: {results_ar}")

Expected Output (structure):

Text (EN): I'm so happy for you, but I'm also a little bit sad to see you go. Predictions: [[{'label': 'joy', 'score': 0.9...}, {'label': 'sadness', 'score': 0.8...}, {'label': 'caring', 'score': 0.5...}, ...]]

Text (ES): ¡Qué sorpresa! No me lo esperaba para nada. Predictions: [[{'label': 'surprise', 'score': 0.9...}, {'label': 'excitement', 'score': 0.4...}, ...]]

Text (AR): أشعر بخيبة أمل وغضب بسبب ما حدث Predictions: [[{'label': 'disappointment', 'score': 0.9...}, {'label': 'anger', 'score': 0.9...}, ...]]

Evaluation

The model's performance was rigorously evaluated on the test set.

Test Set Performance

The following table shows the performance metrics of the fine-tuned model on the test set, broken down by emotion category.

index	accuracy	precision	recall	f1	mcc	support	threshold
admiration	0.942	0.652	0.684	0.667	0.636	2790	0.4
amusement	0.973	0.735	0.817	0.774	0.76	1866	0.35
anger	0.96	0.411	0.364	0.386	0.366	1128	0.35
annoyance	0.896	0.246	0.481	0.325	0.293	1704	0.15
approval	0.91	0.329	0.383	0.354	0.307	2094	0.2
caring	0.958	0.285	0.46	0.352	0.341	816	0.15
confusion	0.965	0.444	0.401	0.421	0.404	1020	0.25
curiosity	0.935	0.433	0.74	0.546	0.535	1734	0.25
desire	0.984	0.404	0.534	0.46	0.457	414	0.25
disappointment	0.942	0.224	0.345	0.272	0.249	1014	0.15
disapproval	0.935	0.306	0.413	0.352	0.322	1398	0.25
disgust	0.975	0.343	0.418	0.377	0.366	600	0.15
embarrassment	0.99	0.28	0.242	0.26	0.255	240	0.1
excitement	0.973	0.344	0.425	0.38	0.369	624	0.15
fear	0.987	0.599	0.522	0.558	0.553	498	0.35
gratitude	0.989	0.924	0.902	0.913	0.907	2004	0.4
grief	0.999	0	0	0	0	36	0.05
joy	0.965	0.454	0.532	0.49	0.474	1032	0.25
love	0.973	0.731	0.829	0.777	0.765	1812	0.35
nervousness	0.996	0.385	0.25	0.303	0.308	120	0.1
optimism	0.973	0.588	0.525	0.555	0.542	1062	0.25
pride	0.997	0	0	0	0	84	0.05
realization	0.962	0.202	0.189	0.195	0.176	792	0.15
relief	0.996	0	0	0	0	138	0.05
remorse	0.988	0.597	0.808	0.687	0.689	516	0.15
sadness	0.97	0.548	0.434	0.484	0.473	1062	0.4
surprise	0.974	0.487	0.569	0.524	0.513	828	0.3
neutral	0.726	0.551	0.818	0.658	0.468	10524	0.2

Use Cases

This model is ideal for applications requiring nuanced emotional understanding across different languages:

Global Customer Feedback Analysis: Analyze customer reviews, support tickets, and survey responses from around the world to gauge sentiment.

Multilingual Social Media Monitoring: Track brand perception and public mood across different regions and languages.

Advanced Chatbot Development: Build more empathetic and responsive chatbots that can understand user emotions in their native language.

Content Moderation: Automatically flag toxic, aggressive, or sensitive content on international platforms.

Market Research: Gain insights into how different cultures express emotions in text.

Trained On

Base Model: google-bert/bert-base-multilingual-cased - A powerful pretrained model supporting 104 languages.

Dataset: multilingual_go_emotions - A carefully translated and curated dataset for multilingual emotion analysis, based on the original Google GoEmotions dataset.

Fine-Tuning Guide

To adapt this model for your own dataset or to replicate the training process, you can follow the methodology outlined in the official code repository. The repository provides a complete, end-to-end example, including data preprocessing, training scripts, and evaluation logic.

For full details, please refer to the GitHub repository: emotion_chatbot

The following plots visualize the model's performance during the fine-tuning process across epochs.

Loss Curves (Training vs. Validation)

Accuracy Curves (Training vs. Validation)

F1 Score Curves (Training vs. Validation)

Tags

#multilingual-nlp #emotion-classification #text-classification #multi-label #bert #transformer #natural-language-processing #sentiment-analysis #deep-learning #arabic-nlp #french-nlp #spanish-nlp #goemotions #BERT-Emotion #edge-nlp #emotion-detection #offline-nlp
#sentiment-analysis #emojis #emotions #embedded-nlp #ai-for-iot #efficient-bert #nlp2025 #context-aware #edge-ml
#smart-home-ai #emotion-aware #voice-ai #eco-ai #chatbot #social-media
#mental-health #short-text #smart-replies #tone-analysis

Support & Contact

For questions, bug reports, or collaboration inquiries, please open an issue on the Hugging Face Hub repository or contact the author directly.

Author: Anas Hamid Alokla

📬 Email: [email protected]