Reproduce “Multi-Target Cross-Lingual Summarization: a novel task and a language-neutral approach” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Diogo Pernes, Gonçalo M. Correia, Afonso Mendes
Researchers from Priberam Labs introduce Multi-Target Cross-Lingual Summarization (MTXLS), a task requiring semantically identical summaries in multiple target languages from a single source document. They propose a language-neutral re-ranking framework, NeutralRR, which consistently achieves higher semantic coherence across target language summaries than pivot-dependent methods, without significantly reducing individual summary quality.
Key insights
- Ensuring semantic coherence across multiple target language summaries presents a distinct challenge beyond individual summary quality in cross-lingual summarization.
- A language-neutral re-ranking approach can effectively promote semantic coherence by directly optimizing for average pairwise similarity across all target summaries.
- Relying on a single pivot language for coherence can introduce biases and limit overall consistency compared to a truly language-neutral strategy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.