2026-10-12T03:22:27Z https://www.scidb.cn/oai

10.57760/sciencedb.j00104.00158 2026-09-10T10:51:56Z

2026-09-10 MDF-DTR: A Multi-Dimensional Feature-Based Dataset for Domain Term Recognition doi:10.57760/sciencedb.j00104.00158 en MDF-DTR is a dataset for domain term recognition in scientific and technical literature. It is constructed from scientific abstracts, author-assigned keywords, and authoritative terminology resources through a multi-dimensional feature-based candidate screening and back-annotation workflow.The dataset includes a candidate term repository, back-annotated training data, manually annotated test data, metadata, and annotation guidelines. The chemistry candidate term repository contains more than 800,000 candidate terms, and the chemistry and biology subsets each contain 100,000 back-annotated instances.MDF-DTR can be used for domain term recognition, terminology extraction, scientific text mining, knowledge organization, and domain-specific information retrieval. Domain Term Recognition; Terminology Extraction; Scientific Text Mining 邵田 RESTRICTED dataset Science Data Bank