LED-Base Book Summary Model
| Property | Value |
|---|---|
| Parameter Count | 162M |
| Model Type | Longformer Encoder-Decoder |
| License | BSD-3-Clause |
| Max Input Length | 16,384 tokens |
| ROUGE-1 Score | 33.45 |
What is led-base-book-summary?
LED-base-book-summary is a specialized text summarization model built on the Longformer Encoder-Decoder architecture. Fine-tuned on the BookSum dataset, this model excels at condensing long-form content while maintaining coherent narrative structure. With 162M parameters, it offers an efficient balance between computational requirements and summary quality.
Implementation Details
The model utilizes a sophisticated architecture that enables processing of up to 16,384 tokens per batch, making it ideal for lengthy documents. It was trained for 16 epochs with carefully optimized learning rates to achieve precise fine-tuning. The model implements special token handling with no-repeat n-gram size of 3 and a repetition penalty of 3.5 to ensure diverse and natural summaries.
- Optimized for long-form technical and narrative content
- Implements SparkNotes-style explanatory summaries
- Supports batch processing of extensive documents
- Features customizable generation parameters
Core Capabilities
- Processes documents up to 16,384 tokens in length
- Generates abstractive summaries with explanatory elements
- Handles technical, academic, and narrative content effectively
- Achieves 33.45 ROUGE-1 score on benchmark tests
Frequently Asked Questions
Q: What makes this model unique?
This model's unique strength lies in its ability to handle extremely long documents while producing coherent, explanatory summaries. It's specifically optimized for book-length and technical content, making it ideal for educational and research applications.
Q: What are the recommended use cases?
The model is best suited for summarizing long narratives, academic papers, textbooks, and technical documents. It's particularly effective when you need to maintain important context while significantly reducing content length.
