Semantic Extractor-Paraphraser based Abstractive Summarization

TL;DR

This paper introduces a semantic overlap-based abstractive summarization system that outperforms existing models and reveals limitations of the Pointer Generator Network in multi-sentence information aggregation.

Contribution

The proposed extractor-paraphraser model emphasizes semantic overlap for summarization, surpassing state-of-the-art baselines and critically analyzing the capabilities of PGN.

Findings

Our model achieves higher ROUGE, METEOR, and WMS scores.

PGN functions more as a paraphraser than a true summarizer.

Extensive ablation experiments validate the effectiveness of the proposed system.

Abstract

The anthology of spoken languages today is inundated with textual information, necessitating the development of automatic summarization models. In this manuscript, we propose an extractor-paraphraser based abstractive summarization system that exploits semantic overlap as opposed to its predecessors that focus more on syntactic information overlap. Our model outperforms the state-of-the-art baselines in terms of ROUGE, METEOR and word mover similarity (WMS), establishing the superiority of the proposed system via extensive ablation experiments. We have also challenged the summarization capabilities of the state of the art Pointer Generator Network (PGN), and through thorough experimentation, shown that PGN is more of a paraphraser, contrary to the prevailing notion of a summarizer; illustrating it's incapability to accumulate information across multiple sentences.