Wav2vec2 Xls R 300m Cv7 Turkish

🚀 wav2vec2-xls-r-300m-cv7-turkish

This ASR model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the Turkish language, offering high - quality automatic speech recognition capabilities for Turkish.

🚀 Quick Start

Before running the evaluation, you need to install the unicode_tr package for Turkish text processing.

Evaluate on mozilla - foundation/common_voice_7_0 with split test

python eval.py --model_id mpoyraz/wav2vec2-xls-r-300m-cv7-turkish --dataset mozilla-foundation/common_voice_7_0 --config tr --split test

Evaluate on speech - recognition-community-v2/dev_data

python eval.py --model_id mpoyraz/wav2vec2-xls-r-300m-cv7-turkish --dataset speech-recognition-community-v2/dev_data --config tr --split validation --chunk_length_s 5.0 --stride_length_s 1.0

✨ Features

• Fine - tuned for Turkish: Based on the pre - trained model facebook/wav2vec2-xls-r-300m, it is fine - tuned on Turkish language data, providing better performance for Turkish speech recognition.
• Multi - dataset support: Supports multiple datasets such as Common Voice 7.0 TR and MediaSpeech.

📦 Installation

There is no specific installation content provided in the original document. If you want to use this model, you may need to install relevant dependencies such as unicode_tr, transformers, pytorch, datasets, and tokenizers according to the framework versions mentioned in the document.

📚 Documentation

Model description

This ASR model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on Turkish language.

Training and evaluation data

The following datasets were used for finetuning:

• Common Voice 7.0 TR: All validated split except test split was used for training.
• MediaSpeech

Training procedure

To support both of the datasets above, custom pre - processing and loading steps were performed and wav2vec2-turkish repo was used for that purpose.

Training hyperparameters

The following hyperparameters were used for finetuning:

• learning_rate: 2e - 4
• num_train_epochs: 10
• warmup_steps: 500
• freeze_feature_extractor
• mask_time_prob: 0.1
• mask_feature_prob: 0.05
• feat_proj_dropout: 0.05
• attention_dropout: 0.05
• final_dropout: 0.05
• activation_dropout: 0.05
• per_device_train_batch_size: 8
• per_device_eval_batch_size: 8
• gradient_accumulation_steps: 8

Framework versions

• Transformers: 4.16.0.dev0
• Pytorch: 1.10.1
• Datasets: 1.17.0
• Tokenizers: 0.10.3

Language Model

An N - gram language model is trained on Turkish Wikipedia articles using KenLM and ngram - lm - wiki repo was used to generate arpa LM and convert it into binary format.

Evaluation results

🔧 Technical Details

The model is based on the pre - trained model facebook/wav2vec2-xls-r-300m and is fine - tuned on Turkish language data. Custom pre - processing and loading steps are used to support multiple datasets. The N - gram language model is trained on Turkish Wikipedia articles to improve the performance of speech recognition.

📄 License

The model is licensed under CC - BY - 4.0.