An approach for automatically structuring Vietnamese legal text

Abstract

Recognizing the structures of legal texts is significant to understand texts in this domain. In this paper, we describe a task of automatically structuring Vietnamese legal documents based on our study of this domain in several aspects: their linguistic features and patterns of recognition with trigger sets. The task focuses on the recognition of the logical structures of the documents stored in law database of Vietnam Ministry of Justice. A rule-based approach in association with some patterns is applied and is verified on a manually built corpus. Experimental result got 64.37% on assumption annotation, 64.15% on provision annotation and 75.76% on sanction annotation in the Fβ=1 score on the corpus of Vietnamese Enterprise Law articles.