Wav2vec2 Large 960h Lv60 Self With Wikipedia Lm

🚀 facebook/wav2vec2-large-960h-lv60-self Enhanced with Wikipedia Language Model

This project enhances the facebook/wav2vec2-large-960h-lv60-self model with a Wikipedia language model, offering improved performance in speech recognition tasks.

🚀 Quick Start

Installation

The installation process is not explicitly provided in the original README. However, it is assumed that you need to have the transformers library installed. You can install it using the following command:

pip install transformers

Usage

from transformers import pipeline
pipe = pipeline("automatic-speech-recognition", model="gxbag/wav2vec2-large-960h-lv60-self-with-wikipedia-lm")
output = pipe("/path/to/audio.wav", chunk_length_s=30, stride_length_s=(6, 3))
print(output)

✨ Features

• Enhanced with Wikipedia Language Model: Leveraging the rich text data from Wikipedia to improve speech recognition accuracy.
• KenLM-based Language Model: A 5-gram model built using KenLM, with pruning of singletons of 3-grams and bigger for efficient performance.

📦 Installation

To use this enhanced model, you need to have the transformers library installed. You can install it via pip:

pip install transformers

💻 Usage Examples

Basic Usage

from transformers import pipeline
pipe = pipeline("automatic-speech-recognition", model="gxbag/wav2vec2-large-960h-lv60-self-with-wikipedia-lm")
output = pipe("/path/to/audio.wav", chunk_length_s=30, stride_length_s=(6, 3))
print(output)

Advanced Usage

There is no advanced usage example provided in the original README. If you have specific advanced scenarios, you can refer to the official documentation of the transformers library.

🔧 Technical Details

Dataset

The dataset used is wikipedia/20200501.en, which includes all articles. The dataset was pre - processed by removing references, external links, and all text inside parentheses. It contains 8092546 words.

Language Model

The language model was built using KenLM. It is a 5-gram model where all singletons of 3-grams and bigger were pruned. The command used to build the model is:

kenlm/build/bin/lmplz -o 5 -S 120G --vocab_estimate 8092546 --text text.txt --arpa text.arpa --prune 0 0 1

📄 License

There is no license information provided in the original README, so this section is skipped.

⚠️ Important Note

In the current version of transformers (as of the release of this model), when using striding in the pipeline, it will chop off the last portion of audio (in this case, 3 seconds). As a workaround, add 3 seconds of silence to the end of the audio. This problem was fixed in the GitHub version of transformers.