wav2vec2-base Arabic Speech Emotion Recognition Model - Open-source and High-precision Recognition of Speech Emotions

Wav2vec2 Base Arabic Speech Emotion Recognition

Developed by ahmmedasaad2772

A fine-tuned Arabic speech emotion recognition model based on facebook/wav2vec2-base, achieving 99.92% accuracy on the evaluation dataset.

Audio Classification

Transformers

Open Source License:Apache-2.0 #Arabic Speech Analysis #High-precision Emotion Recognition #wav2vec2 Fine-tuning

Downloads 352

Release Time : 3/5/2024

Model Overview

This model is specifically designed for recognizing emotions in Arabic speech, fine-tuned based on the wav2vec2 architecture, suitable for speech emotion analysis tasks.

Model Features

High Accuracy

Achieved 99.92% accuracy on the evaluation dataset, demonstrating excellent performance.

Arabic-specific

Optimized specifically for Arabic speech, suitable for Arabic emotion recognition scenarios.

Based on wav2vec2

Utilizes the mature wav2vec2 architecture as its foundation, featuring powerful speech feature extraction capabilities.

Model Capabilities

Arabic Speech Emotion Recognition

Speech Emotion Classification

Speech Feature Extraction

Use Cases

Emotion Analysis

Customer Service Emotion Analysis

Analyze customer emotional states in call center conversations

Accurately identifies emotional states such as anger and satisfaction

Voice Assistant Emotional Response

Adjust voice assistant responses based on user speech emotions

Enhances user experience and interaction naturalness

🚀 wav2vec2-base-arabic_speech_emotion_recognition

This model is a fine - tuned version of facebook/wav2vec2-base on the audiofolder dataset, achieving high accuracy in speech emotion recognition.

🚀 Quick Start

This model is a fine - tuned version of facebook/wav2vec2-base on the audiofolder dataset. It achieves the following results on the evaluation set:

Loss: 0.0430
Accuracy: 0.9992
Weighted f1: 0.9992
Micro f1: 0.9992
Macro f1: 0.9993
Weighted recall: 0.9992
Micro recall: 0.9992
Macro recall: 0.9992
Weighted precision: 0.9992
Micro precision: 0.9992
Macro precision: 0.9993

📚 Documentation

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 3e - 05
train_batch_size: 32
eval_batch_size: 32
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Weighted f1	Micro f1	Macro f1	Weighted recall	Micro recall	Macro recall	Weighted precision	Micro precision	Macro precision
1.6078	0.98	29	1.4403	0.5227	0.4478	0.5227	0.4426	0.5227	0.5227	0.5187	0.4200	0.5227	0.4145
1.6078	1.98	58	0.7310	0.9808	0.9808	0.9808	0.9809	0.9808	0.9808	0.9799	0.9816	0.9808	0.9827
1.6078	2.98	87	0.2797	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993
1.6078	3.98	116	0.1361	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993
1.6078	4.98	145	0.0855	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993
1.6078	5.98	174	0.0638	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993
1.6078	6.98	203	0.0529	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993
1.6078	7.98	232	0.0470	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993
1.6078	8.98	261	0.0440	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993
1.6078	9.98	290	0.0430	0.9992	0.9992	0.9992	0.9993	0.9992	0.9992	0.9992	0.9992	0.9992	0.9993

Framework versions

Transformers 4.26.1
Pytorch 2.1.0+cu121
Datasets 2.18.0
Tokenizers 0.13.3

📄 License

This model is licensed under the Apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご