CroCoSum: A Benchmark Dataset for Cross-Lingual Code-Switched Summarization

TL;DR

CroCoSum is a new dataset for cross-lingual code-switched summarization, capturing natural language phenomena like code-switching in multilingual news summaries, enabling better research in this underexplored area.

Contribution

The paper introduces CroCoSum, a large-scale dataset of real-world code-switched summaries, and evaluates existing models, highlighting challenges and limitations in current approaches.

Findings

Existing CLS resources do not improve CroCoSum performance.

Most summaries contain code-switching, reflecting natural multilingual communication.

Current models struggle with code-switched summarization, indicating need for specialized methods.

Abstract

Cross-lingual summarization (CLS) has attracted increasing interest in recent years due to the availability of large-scale web-mined datasets and the advancements of multilingual language models. However, given the rareness of naturally occurring CLS resources, the majority of datasets are forced to rely on translation which can contain overly literal artifacts. This restricts our ability to observe naturally occurring CLS pairs that capture organic diction, including instances of code-switching. This alteration between languages in mid-message is a common phenomenon in multilingual settings yet has been largely overlooked in cross-lingual contexts due to data scarcity. To address this gap, we introduce CroCoSum, a dataset of cross-lingual code-switched summarization of technology news. It consists of over 24,000 English source articles and 18,000 human-written Chinese news summaries, with more than 92% of the summaries containing code-switched phrases. For reference, we evaluate the performance of existing approaches including pipeline, end-to-end, and zero-shot methods. We show that leveraging existing CLS resources as a pretraining step does not improve performance on CroCoSum, indicating the limited generalizability of current datasets. Finally, we discuss the challenges of evaluating cross-lingual summarizers on code-switched generation through qualitative error analyses.