Extractive Opinion Summarization in Quantized Transformer Spaces

TL;DR

This paper introduces QT, an unsupervised extractive opinion summarization system based on quantized transformer spaces, capable of discovering popular opinions and generating aspect-specific summaries without additional training.

Contribution

The paper proposes QT, a novel unsupervised method using vector quantization for opinion summarization, and introduces SPACE, a new large-scale benchmark for evaluation.

Findings

QT outperforms competitive baselines in human evaluations.

The system can generate aspect-specific summaries without retraining.

SPACE provides a comprehensive benchmark for opinion summarization.

Abstract

We present the Quantized Transformer (QT), an unsupervised system for extractive opinion summarization. QT is inspired by Vector-Quantized Variational Autoencoders, which we repurpose for popularity-driven summarization. It uses a clustering interpretation of the quantized space and a novel extraction algorithm to discover popular opinions among hundreds of reviews, a significant step towards opinion summarization of practical scope. In addition, QT enables controllable summarization without further training, by utilizing properties of the quantized space to extract aspect-specific summaries. We also make publicly available SPACE, a large-scale evaluation benchmark for opinion summarizers, comprising general and aspect-specific summaries for 50 hotels. Experiments demonstrate the promise of our approach, which is validated by human studies where judges showed clear preference for our method over competitive baselines.