distilbart-tos-summarizer-tosdr
| Property | Value |
|---|---|
| Base Model | DistilBART 6-6 |
| Task | Terms of Service Summarization |
| Language | English |
| Framework | PyTorch |
What is distilbart-tos-summarizer-tosdr?
This is a specialized model designed to summarize Terms of Service (ToS) documents, developed by ML6 team. It implements a hybrid approach combining LSA (Latent Semantic Analysis) extractive summarization with abstractive summarization using DistilBART architecture. The model was fine-tuned on TOSDR (Terms of Service Didn't Read) dataset to specifically handle legal and technical content found in terms of service documents.
Implementation Details
The model utilizes a two-stage summarization pipeline: first applying extractive summarization to reduce the input text, followed by abstractive summarization using the fine-tuned DistilBART model. It supports maximum input length of 1024 tokens and can generate summaries between 5 and 512 tokens.
- Built on sshleifer/distilbart-cnn-6-6 architecture
- Incorporates LSA extractive summarization as preprocessing
- Optimized for Terms of Service content
- Supports batch processing and inference endpoints
Core Capabilities
- Efficient processing of long legal documents
- Hybrid extractive-abstractive summarization
- Specialized in understanding legal terminology
- Maintains semantic accuracy in summaries
Frequently Asked Questions
Q: What makes this model unique?
The model's unique strength lies in its specialized focus on Terms of Service documents and its hybrid summarization approach, combining both extractive and abstractive methods for optimal results.
Q: What are the recommended use cases?
This model is ideal for automatically summarizing Terms of Service documents, privacy policies, and other legal documents to make them more accessible to users. It's particularly useful for companies wanting to provide clear, concise versions of their legal documents.
