metadata

license: mit
language:
  - en
metrics:
  - rouge
base_model:
  - allenai/led-base-16384
pipeline_tag: summarization
library_name: transformers

Longformer fine-tuned to summarize Terms of Service

Terms of Service documents are lengthy, complex, and time-consuming to read. Due to its vague language, people often don’t understand what they are agreeing to. Hence, we have fine-tuned Longformer model to help summarize TOS and make it easy to read and understand.

This model is a fine-tuned version of allenai/led-base-16384
Dataset used: TL;DRLegal and TOS;DR website
It achieves the following results on the validation set:

ROUGE-1: 0.28
ROUGE-2: 0.13
ROUGE-L: 0.27

How to Use

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("aarushi-211/TOS-Longformer")
model = AutoModelForSequenceClassification.from_pretrained("aarushi-211/TOS-Longformer")

inputs = tokenizer("Your input text here", return_tensors="pt", truncation=True)
outputs = model(**inputs)
logits = outputs.logits