Retrieval-Augmented Multilingual Keyphrase Generation with Retriever-Generator Iterative Training

TL;DR

This paper introduces a retrieval-augmented multilingual keyphrase generation method that leverages English datasets and cross-lingual retrieval to improve keyphrase prediction in low-resource languages, supported by new datasets and iterative training.

Contribution

It proposes a novel retrieval-augmented approach with iterative training for multilingual keyphrase generation and introduces two new multilingual datasets.

Findings

Outperforms all baseline methods in experiments.

Effectively leverages English keyphrase data for low-resource languages.

Improves cross-lingual keyphrase generation accuracy.

Abstract

Keyphrase generation is the task of automatically predicting keyphrases given a piece of long text. Despite its recent flourishing, keyphrase generation on non-English languages haven't been vastly investigated. In this paper, we call attention to a new setting named multilingual keyphrase generation and we contribute two new datasets, EcommerceMKP and AcademicMKP, covering six languages. Technically, we propose a retrieval-augmented method for multilingual keyphrase generation to mitigate the data shortage problem in non-English languages. The retrieval-augmented model leverages keyphrase annotations in English datasets to facilitate generating keyphrases in low-resource languages. Given a non-English passage, a cross-lingual dense passage retrieval module finds relevant English passages. Then the associated English keyphrases serve as external knowledge for keyphrase generation in the current language. Moreover, we develop a retriever-generator iterative training algorithm to mine pseudo parallel passage pairs to strengthen the cross-lingual passage retriever. Comprehensive experiments and ablations show that the proposed approach outperforms all baselines.