Using Multimodal and Language-Agnostic Sentence Embeddings for Abstractive Summarization

TL;DR

This paper introduces SBARThez, a framework that uses multimodal, language-agnostic sentence embeddings and a Named Entity Injection mechanism to improve the factual accuracy and cross-lingual capabilities of abstractive summarization models.

Contribution

It proposes a novel framework combining pretrained multimodal embeddings and entity injection to enhance factual consistency and multilingual summarization in abstractive models.

Findings

Competitive performance on low-resource languages

Improved factual accuracy with Named Entity Injection

Supports both text and speech inputs

Abstract

Abstractive summarization aims to generate concise summaries by creating new sentences, allowing for flexible rephrasing. However, this approach can be vulnerable to inaccuracies, particularly `hallucinations' where the model introduces non-existent information. In this paper, we leverage the use of multimodal and multilingual sentence embeddings derived from pretrained models such as LaBSE, SONAR, and BGE-M3, and feed them into a modified BART-based French model. A Named Entity Injection mechanism that appends tokenized named entities to the decoder input is introduced, in order to improve the factual consistency of the generated summary. Our novel framework, SBARThez, is applicable to both text and speech inputs and supports cross-lingual summarization; it shows competitive performance relative to token-level baselines, especially for low-resource languages, while generating more concise and abstract summaries.