🔍Automatic Subtitle Generation for Bengali Multimedia Using Deep Learning📊🔬

In this repository, you'll find the code and resources for our project focused on automating Bengali subtitle generation for multimedia content. We've harnessed the power of deep learning techniques and Automatic Speech Recognition (ASR) to create a system that accurately transcribes Bengali audio into text, synchronizing it seamlessly with multimedia.

🌐 Our system utilizes the state-of-the-art Wav2Vec2 model, fine-tuned on the Common Voice Bengali Dataset, and employs linguistic considerations specific to the Bengali language for improved accuracy. We've achieved impressive results in Character Error Rate (CER) and Word Error Rate (WER) and ensured precise timecode generation for subtitles.

🧐 Key Features:

• Training and fine-tuning of the Wav2Vec2 ASR model for the Bengali language.
• Subtitle generation system with Timecode synchronization.
• Low Character Error Rate (CER), Word Error Rate (WER), and High Timecode Accuracy for precise transcriptions.
• Comparison between Manually-Created vs Machine-Generated subtitles.
• Accurate distribution of the number of words in the sentences.

Publication

• Published in: []
• DOI: []