Wav2vec2-large-960h-intent-classification-ori Open-source Model - A Powerful Tool for High-accuracy Speech Intent Classification

Wav2vec2 Large 960h Intent Classification Ori

Developed by MuhammadIqbalBazmi

Fine-tuned intent classification model based on facebook/wav2vec2-large-960h, achieving 77.08% accuracy on the evaluation set

Audio Classification

Transformers

Open Source License:Apache-2.0 #Speech Intent Classification #wav2vec2 Fine-tuning #High Accuracy

Downloads 15

Release Time : 10/11/2022

Model Overview

This model is a speech intent classification model fine-tuned based on the wav2vec2-large-960h architecture, primarily used to recognize user intent from speech

Model Features

High Accuracy

Achieves 77.08% accuracy on the evaluation set

Based on Pre-trained Model

Fine-tuned from facebook/wav2vec2-large-960h, with powerful speech feature extraction capabilities

Linear Learning Rate Scheduling

Uses linear learning rate scheduling strategy with warm-up mechanism to optimize the training process

Model Capabilities

Speech Intent Recognition

Speech Classification

Speech Feature Extraction

Use Cases

Virtual Assistant

Voice Command Recognition

Recognizes the intent of user voice commands

77.08% accuracy

Customer Service System

Automatic Voice Response

Automatically classifies and routes services based on user speech content

🚀 wav2vec2-large-960h-intent-classification-ori

This model is a fine - tuned version of facebook/wav2vec2-large-960h, which can be used for intent classification.

🚀 Quick Start

This model is a fine-tuned version of facebook/wav2vec2-large-960h on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.6013
Accuracy: 0.7708

📚 Documentation

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 3e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 45

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.1937	1.0	14	2.1715	0.3333
2.1764	2.0	28	2.1317	0.3333
2.1358	3.0	42	2.0794	0.3333
2.1227	4.0	56	2.0367	0.3333
2.1157	5.0	70	1.9845	0.3333
2.0776	6.0	84	1.9369	0.3333
2.0126	7.0	98	1.8021	0.375
1.88	8.0	112	1.6825	0.4167
1.8818	9.0	126	1.4638	0.5
1.7044	10.0	140	1.4188	0.5208
1.5442	11.0	154	1.3421	0.5625
1.5045	12.0	168	1.3971	0.5
1.3369	13.0	182	1.1602	0.5833
1.4017	14.0	196	1.3510	0.5417
1.2565	15.0	210	1.0978	0.5625
1.1056	16.0	224	1.0847	0.5833
1.2006	17.0	238	1.0262	0.625
0.9235	18.0	252	0.9532	0.7083
0.9528	19.0	266	1.0212	0.6042
0.8195	20.0	280	0.8442	0.7083
0.7518	21.0	294	0.8379	0.6875
0.6017	22.0	308	0.9422	0.7292
0.7697	23.0	322	0.7353	0.75
0.5367	24.0	336	0.8685	0.6875
0.5655	25.0	350	0.7440	0.7708
0.5116	26.0	364	0.7572	0.75
0.4297	27.0	378	0.7518	0.75
0.4928	28.0	392	0.5988	0.7917
0.4424	29.0	406	0.7240	0.75
0.3313	30.0	420	0.6173	0.7708
0.3854	31.0	434	0.7375	0.75
0.4131	32.0	448	0.7026	0.7708
0.2899	33.0	462	0.6516	0.7708
0.3644	34.0	476	0.6201	0.7917
0.2316	35.0	490	0.6111	0.7708
0.2589	36.0	504	0.5518	0.7917
0.3778	37.0	518	0.5512	0.7708
0.2426	38.0	532	0.5779	0.7917
0.304	39.0	546	0.7771	0.75
0.1833	40.0	560	0.5839	0.7708
0.1649	41.0	574	0.5699	0.7708
0.2529	42.0	588	0.6190	0.75
0.2121	43.0	602	0.5992	0.75
0.2736	44.0	616	0.6011	0.7917
0.2446	45.0	630	0.6013	0.7708

Framework versions

Transformers 4.20.1
Pytorch 1.11.0
Datasets 2.1.0
Tokenizers 0.12.1

📄 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