muaalem-model-v3

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0118
• Per Phonemes: 0.0043
• Per Hams Or Jahr: 0.0020
• Per Shidda Or Rakhawa: 0.0027
• Per Tafkheem Or Taqeeq: 0.0031
• Per Itbaq: 0.0013
• Per Safeer: 0.0014
• Per Qalqla: 0.0013
• Per Tikraar: 0.0037
• Per Tafashie: 0.0019
• Per Istitala: 0.0012
• Per Ghonna: 0.0017
• Average Per: 0.0022

Model description

The model was presented in the paper Automatic Pronunciation Error Detection and Correction of the Holy Quran's Learners Using Deep Learning

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Per Phonemes	Per Hams Or Jahr	Per Shidda Or Rakhawa	Per Tafkheem Or Taqeeq	Per Itbaq	Per Safeer	Per Qalqla	Per Tikraar	Per Tafashie	Per Istitala	Per Ghonna	Average Per
0.128	0.2002	650	0.0237	0.0075	0.0031	0.0043	0.0071	0.0020	0.0020	0.0019	0.0056	0.0037	0.0018	0.0024	0.0038
0.0172	0.4005	1300	0.0128	0.0038	0.0017	0.0025	0.0039	0.0013	0.0013	0.0012	0.0044	0.0024	0.0011	0.0016	0.0023
0.0146	0.6007	1950	0.0105	0.0033	0.0014	0.0022	0.0028	0.0010	0.0012	0.0011	0.0039	0.0017	0.0009	0.0015	0.0019
0.0111	0.8010	2600	0.0118	0.0043	0.0020	0.0027	0.0031	0.0013	0.0014	0.0013	0.0037	0.0019	0.0012	0.0017	0.0022

Framework versions

• Transformers 4.55.4
• Pytorch 2.8.0+cu128
• Datasets 3.3.2
• Tokenizers 0.21.4