Here, I have summarized YouTube Video Transcripts & Blog posts with SOTA models. The Video Summarization was done through the facebook/bart-large-cnn model from HuggingFace Transformers. With the sshleifer/distilbart-cnn model, we fetched and scraped a blog post to return its summary to the user. Lastly, the latest Pegasus model from Google was also run to generate an Abstractive summary of the text from the corpus given, which can generate text from its own & doesn't use the text from the corpus! All these approaches use current technologies in Text Summarization applying recent breakthroughs.
Extractive Summary models (Bart) was able to reduce the size of the text corpus to 20-30%, while still retaining the meaning of the corpus. The Abstractive Summary model (Pegasus) often returned only a couple of sentences even after providing a giant corpus.
- Python
- 🤗 Transformers
- Youtube Transcript API
- BeautifulSoup
- Tensorflow
- PyTorch
- YouTube Video Transcription & Summarisation with the facebook/bart-large-cnn model.
- Blog Post Summarization with the default sshleifer/distilbart-cnn model from HuggingFace library.
- Abstractive Summary of a text corpus with the latest Pegasus model from Google.
