Leveraging Digitized Newspapers to Collect Summarization Data in Low-Resource Languages

TL;DR

This paper introduces a novel method to extract summarization data from digitized newspapers' front-page teasers across multiple languages, enabling the creation of new datasets for low-resource languages like Hebrew.

Contribution

The work presents a scalable automatic process to collect naturally occurring summaries from digitized newspapers, supporting multi-document summarization in under-represented languages.

Findings

Method successfully applied to seven languages.

Created HEBTEASESUM, the first Hebrew multi-document summarization dataset.

Demonstrated the method's scalability and language independence.

Abstract

High quality summarization data remains scarce in under-represented languages. However, historical newspapers, made available through recent digitization efforts, offer an abundant source of untapped, naturally annotated data. In this work, we present a novel method for collecting naturally occurring summaries via Front-Page Teasers, where editors summarize full length articles. We show that this phenomenon is common across seven diverse languages and supports multi-document summarization. To scale data collection, we develop an automatic process, suited to varying linguistic resource levels. Finally, we apply this process to a Hebrew newspaper title, producing HEBTEASESUM, the first dedicated multi-document summarization dataset in Hebrew.