Rubert Tiny2 Cedr Emotion Detection

🚀 Fine-tuned Model for Russian Sentence Emotion Classification

This project presents a fine - tuned model based on cointegrated/rubert-tiny2 for classifying emotions in Russian sentences. It addresses the multilabel classification task, as a single sentence may convey multiple emotions.

🚀 Quick Start

This is the fine - tuned cointegrated/rubert-tiny2 model for classifying emotions in Russian sentences. The task is multilabel classification since one sentence can contain multiple emotions.

✨ Features

• Multilabel Classification: Capable of identifying multiple emotions within a single Russian sentence.
• Trained on CEDR Dataset: Utilizes the CEDR dataset described in the paper "Data - Driven Model for Emotion Detection in Russian Texts" by Sboev et al.

🔧 Technical Details

The model was trained with the Adam optimizer for 40 epochs. The learning rate was set to 1e - 5, and the batch size was 64. You can find the training details in this notebook.

📚 Documentation

Model Performance

The quality of the predicted probabilities on the test dataset is as follows:

Additional Information

• Language: Russian
• Tags: russian, classification, sentiment, emotion - classification, multiclass
• Datasets: cedr

You can test the model with the following sample texts:

• "Бесишь меня, падла"
• "Как здорово, что все мы здесь сегодня собрались"
• "Как-то стрёмно, давай свалим отсюда?"
• "Грусть-тоска меня съедает"
• "Данный фрагмент текста не содержит абсолютно никаких эмоций"
• "Нифига себе, неужели так тоже бывает!"