Title: Multi-stage Information Retrieval for Vietnamese Legal Texts

Abstract: This study deals with the problem of information retrieval (IR) for Vietnamese legal texts. Despite being well researched in many languages, information retrieval has still not received much attention from the Vietnamese research community. This is especially true for the case of legal documents, which are hard to process. This study proposes a new approach for information retrieval for Vietnamese legal documents using sentence-transformer. Besides, various experiments are conducted to make comparisons between different transformer models, ranking scores, syllable-level, and word-level training. The experiment results show that the proposed model outperforms models used in current research on information retrieval for Vietnamese documents.