Unsupervised Opinion Summarization with Content Planning

TL;DR

This paper introduces a content planning approach for unsupervised opinion summarization that improves summary quality and enables the creation of more natural synthetic datasets without needing expensive annotations.

Contribution

It proposes a novel content planning method using aspect and sentiment distributions, enhancing unsupervised summarization and synthetic data generation.

Findings

Outperforms existing models in summary informativeness and coherence

Generates summaries that better capture opinion consensus

Creates more realistic synthetic datasets for training

Abstract

The recent success of deep learning techniques for abstractive summarization is predicated on the availability of large-scale datasets. When summarizing reviews (e.g., for products or movies), such training data is neither available nor can be easily sourced, motivating the development of methods which rely on synthetic datasets for supervised training. We show that explicitly incorporating content planning in a summarization model not only yields output of higher quality, but also allows the creation of synthetic datasets which are more natural, resembling real world document-summary pairs. Our content plans take the form of aspect and sentiment distributions which we induce from data without access to expensive annotations. Synthetic datasets are created by sampling pseudo-reviews from a Dirichlet distribution parametrized by our content planner, while our model generates summaries based on input reviews and induced content plans. Experimental results on three domains show that our approach outperforms competitive models in generating informative, coherent, and fluent summaries that capture opinion consensus.