MuLMS

MuLMS是一个包含50篇开放获取材料科学领域文章的大型数据集，由博世人工智能中心等机构创建。该数据集通过领域专家进行多层标注，包括命名实体、关系和框架结构，旨在支持信息抽取任务。数据集涵盖了材料科学的七个子领域，通过多任务训练与现有资源结合，展示了其在提高模型性能方面的潜力。MuLMS不仅支持传统的信息抽取任务，还促进了新型语义搜索应用的发展，如通过特定条件搜索文献中的测量报告。

MuLMS is a large-scale dataset consisting of 50 open-access articles in the field of materials science, developed by institutions including the Bosch Center for Artificial Intelligence and other relevant organizations. This dataset features multi-layer annotations conducted by domain experts, covering named entities, relationships and frame structures, and is designed to support information extraction tasks. Covering seven subfields of materials science, the dataset combines multi-task training with existing resources and demonstrates its potential to enhance model performance. Moreover, MuLMS not only supports traditional information extraction tasks but also facilitates the development of novel semantic search applications, such as retrieving measurement reports in academic literature based on specific criteria.