wav2vec_base_crema_sentiment_analysis - Open-source Speech Emotion Analysis Model with 70.87% Accuracy

Wav2vec Base Crema Sentiment Analysis

Developed by Piyush2512

A speech emotion analysis model fine-tuned based on facebook/wav2vec2-base, achieving 70.87% accuracy on the evaluation set

Audio Classification

Transformers

Open Source License:Apache-2.0 #Speech Emotion Analysis #High Accuracy Top3 #wav2vec2 Fine-tuning

Downloads 38

Release Time : 4/23/2024

Model Overview

This model is a deep learning model for emotion analysis of speech signals, capable of identifying emotional categories from speech. Based on the wav2vec2 architecture, it is suitable for speech emotion classification tasks.

Model Features

High Top3 Accuracy

Achieves 94.5% Top3 accuracy on the evaluation set, effectively identifying major emotional categories

Based on wav2vec2 Architecture

Utilizes the powerful speech feature extraction capability of wav2vec2 to achieve high-quality emotion analysis

Fine-tuning

After 30 epochs of fine-tuning, the model performance has been significantly improved

Model Capabilities

Speech Emotion Recognition

Emotion Classification

Speech Feature Extraction

Use Cases

Emotion Analysis

Customer Service Call Analysis

Analyze the emotional tendencies in customer calls

Can identify customer satisfaction

Mental Health Assessment

Analyze user emotional states through speech

Assists in mental health monitoring

🚀 wav2vec_base_crema_sentiment_analysis

This model is a fine - tuned version of facebook/wav2vec2-base, aiming to perform sentiment analysis on audio data.

🚀 Quick Start

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

Top3 Accuracy: 0.9450
Loss: 0.8840
Accuracy: 0.7087

📚 Documentation

Training and Evaluation

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e - 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: 30

Training results

Training Loss	Epoch	Step	Top3 Accuracy	Validation Loss	Accuracy
1.7821	0.9829	43	0.6711	1.7729	0.2957
1.6948	1.9886	87	0.7643	1.6510	0.3513
1.4876	2.9943	131	0.8593	1.4541	0.4615
1.3657	4.0	175	0.9095	1.3136	0.5421
1.2277	4.9829	218	0.9220	1.2109	0.5824
1.1083	5.9886	262	0.9203	1.1539	0.6030
1.0069	6.9943	306	0.9382	1.0568	0.6496
0.9566	8.0	350	0.9337	1.0153	0.6667
0.8801	8.9829	393	0.9373	0.9970	0.6622
0.8529	9.9886	437	0.9346	0.9792	0.6720
0.7565	10.9943	481	0.9471	0.9475	0.6882
0.7427	12.0	525	0.9462	0.9413	0.6783
0.6616	12.9829	568	0.9516	0.9155	0.6980
0.6539	13.9886	612	0.9543	0.9015	0.6944
0.6036	14.9943	656	0.9471	0.8954	0.6962
0.607	16.0	700	0.9507	0.9088	0.7007
0.5829	16.9829	743	0.9471	0.8934	0.7043
0.5772	17.9886	787	0.9543	0.9182	0.6837
0.5332	18.9943	831	0.9552	0.8802	0.7052
0.5096	20.0	875	0.9525	0.9697	0.6676
0.524	20.9829	918	0.9588	0.8813	0.7061
0.5195	21.9886	962	0.9588	0.8753	0.7142
0.4594	22.9943	1006	0.9552	0.9003	0.7007
0.4478	24.0	1050	0.9561	0.8869	0.6998
0.4578	24.9829	1093	0.9624	0.8874	0.7070
0.4516	25.9886	1137	0.9606	0.8648	0.7142
0.4574	26.9943	1181	0.9597	0.8755	0.7133
0.4093	28.0	1225	0.9615	0.8804	0.7043
0.4216	28.9829	1268	0.9606	0.8814	0.7088
0.4257	29.4857	1290	0.9606	0.8805	0.7097

Framework versions

Transformers 4.40.0
Pytorch 2.2.1+cu121
Datasets 2.19.0
Tokenizers 0.19.1

📄 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