# STRASS: A Light and Effective Method for Extractive Summarization Based on Sentence Embeddings

## TL;DR

STRASS is a computationally efficient extractive summarization method that uses sentence embeddings and a learned transformation to select sentences closest to the document embedding, achieving competitive results.

## Contribution

The paper introduces STRASS, a novel extractive summarization approach that leverages sentence embeddings and a simple transformation, enabling fast training and inference.

## Key findings

- Performs similarly to state-of-the-art methods on benchmark datasets.
- Training and inference are inexpensive and fast due to the simple model.
- Introduces the French CASS dataset for summarization evaluation.

## Abstract

This paper introduces STRASS: Summarization by TRAnsformation Selection and Scoring. It is an extractive text summarization method which leverages the semantic information in existing sentence embedding spaces. Our method creates an extractive summary by selecting the sentences with the closest embeddings to the document embedding. The model learns a transformation of the document embedding to minimize the similarity between the extractive summary and the ground truth summary. As the transformation is only composed of a dense layer, the training can be done on CPU, therefore, inexpensive. Moreover, inference time is short and linear according to the number of sentences. As a second contribution, we introduce the French CASS dataset, composed of judgments from the French Court of cassation and their corresponding summaries. On this dataset, our results show that our method performs similarly to the state of the art extractive methods with effective training and inferring time.