Fine-Tuned Pegasus Transcript Summarizer
Model Details
Model Description
This model is a fine-tuned version of google/pegasus-large optimized for text summarization, specifically designed for generating concise and structured summaries from spoken transcripts. It captures key takeaways, removes filler content, and retains core semantic coherence.
- Developed by: Sandeep Ganga
- Model type: Transformer-based sequence-to-sequence summarization model
- Language(s): English
- License: Apache 2.0
- Finetuned from model: google/pegasus-large
Model Sources
Uses
Direct Use
- Meeting and lecture transcript summarization
- Podcast and interview summaries
- Long-form unstructured text condensor
Downstream Use
The model can be fine-tuned further for:
- Domain-specific summarization (e.g., medical, legal, educational transcripts)
- Integration into AI-powered note-taking tools
Out-of-Scope Use
- Generating fictional or creative stories
- Processing highly degraded or unsegmented audio transcripts
- Producing legally binding or medical summaries without human review
Bias, Risks, and Limitations
- Dataset Bias: The model may reflect domain biases present in the fine-tuning transcript dataset.
- Loss of Context: Highly complex or lengthy transcripts may experience minor context compression.
- Accuracy: Generated output should be validated for mission-critical applications to avoid hallucinations.
How to Get Started with the Model
Use the code below to run inference with PyTorch:
from transformers import PegasusForConditionalGeneration, PegasusTokenizer
model_id = "sandeepsaga/Transcript_Summerizer"
tokenizer = PegasusTokenizer.from_pretrained(model_id)
model = PegasusForConditionalGeneration.from_pretrained(model_id)
def summarize_text(text):
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512, padding="longest")
summary_ids = model.generate(**inputs, max_length=150, min_length=30, length_penalty=2.0)
return tokenizer.decode(summary_ids[0], skip_special_tokens=True)
# Example usage
sample_transcript = "The team met to review quarterly progress. Overall performance met expectations..."
print(summarize_text(sample_transcript))