Retrieval-Augmented Multilingual Keyphrase Generation with Retriever-Generator Iterative Training
This addresses the data shortage problem for keyphrase generation in non-English languages, though it is incremental as it adapts existing retrieval-augmented techniques to a new multilingual setting.
The paper tackled multilingual keyphrase generation by proposing a retrieval-augmented method that uses English keyphrases to aid generation in low-resource languages, and it outperformed all baselines in experiments.
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.