MEDDISTANT19

MEDDISTANT19是一个用于广泛覆盖生物医学关系抽取的准确基准数据集，由德国人工智能研究中心创建。该数据集通过将MEDLINE摘要与广泛使用的SNOMED临床术语知识库对齐获得，旨在解决现有基准数据集中的训练和测试关系重叠问题。MEDDISTANT19包含22种关系，覆盖了大量的生物医学概念及其语义类型，适用于大规模生物医学关系抽取的研究。数据集的创建过程涉及使用SCISPACY进行句子标记化和实体链接，确保了数据的质量和适用性。该数据集的应用领域包括生物分子信息抽取、药物基因组学和药物相互作用识别等，旨在提高生物医学领域的知识发现和管理效率。MEDDISTANT19 is an accurate benchmark dataset designed for comprehensive biomedical relation extraction, developed by the German Research Center for Artificial Intelligence (DFKI). It is constructed by aligning MEDLINE abstracts with the widely adopted SNOMED Clinical Terms knowledge base, with the goal of resolving the training-test relation overlap problem present in existing benchmark datasets. MEDDISTANT19 encompasses 22 distinct relation types, covers a broad range of biomedical concepts and their semantic types, and is suitable for large-scale biomedical relation extraction research. The dataset construction process utilizes SCISPACY for sentence tokenization and entity linking, which guarantees the quality and applicability of the dataset. Application scenarios of this dataset include biomolecular information extraction, pharmacogenomics, drug interaction recognition and other related fields, aiming to enhance the efficiency of knowledge discovery and management in the biomedical domain.

搜集汇总

背景与挑战

背景概述

MEDDISTANT19是一个由德国人工智能研究中心创建的生物医学关系抽取基准数据集，通过MEDLINE摘要与SNOMED临床术语知识库对齐构建，解决了训练和测试关系重叠问题。该数据集包含22种关系，覆盖大量生物医学概念，适用于生物分子信息抽取、药物基因组学等应用，旨在提高知识发现效率。