Spec Soul Ast

🚀 Finetuned Audio Spectrogram Transformer for Russian Sentiment Analysis

This project presents a finetuned Audio Spectrogram Transformer designed for sentiment analysis in Russian. It offers a practical solution for understanding emotional tones in Russian audio data.

🚀 Quick Start

The model is based on the Audio Spectrogram Transformer, finetuned for sentiment analysis in Russian. You can access the code and a Telegram bot implementation in the GitHub repository.

📚 Documentation

Dataset

• Aniemore/resd

Metrics

• Accuracy

Library

• transformers

Pipeline

• audio-classification

Initial State

The model starts from the state MIT/ast-finetuned-audioset-10-10-0.4593.

Model Performance

              precision    recall  f1-score   support

           0       0.77      0.77      0.77        44
           1       0.54      0.59      0.56        37
           2       0.53      0.60      0.56        40
           3       0.69      0.64      0.67        45
           4       0.56      0.57      0.56        44
           5       0.49      0.55      0.52        38
           6       0.75      0.47      0.58        32

    accuracy                           0.61       280
    macro avg      0.62      0.60      0.60       280
    weighted avg   0.62      0.61      0.61       280