We propose a domain adaptation method for multilingual sentence encoders. In domains requiring a high level of expertise, such as medical and academic, domain-specific pre-trained models have been released in each language. However, there is no its multilingual version, which prevents application to cross-lingual information retrieval. Obviously, multilingual pre-training with developing in-domain corpora in each language is costly. Therefore, we efficiently develop domain-specific cross-lingual sentence encoders from existing multilingual sentence encoders and domain-specific monolingual sentence encoders in each language. Experimental results on translation ranking in three language pairs with different domains reveal the effectiveness of the proposed method compared to baselines without domain adaptation and existing domain adaptation methods.

We release a pretrained Japanese masked language model for an academic domain. Pretrained masked language models have recently improved the performance of various natural language processing applications. In domains such as medical and academic, which include a lot of technical terms, domain-specific pretraining is effective. While domain-specific masked language models for medical and SNS domains are widely used in Japanese, along with domain-independent ones, pretrained models specific to the academic domain are not publicly available. In this study, we pretrained a RoBERTa-based Japanese masked language model on paper abstracts from the academic database CiNii Articles. Experimental results on Japanese text classification in the academic domain revealed the effectiveness of the proposed model over existing pretrained models.