NIMS polymer database PoLyInfo (III): modularizing ShEx schemas for descriptors and properties in PoLyInfoRDF

PoLyInfo is a polymer database of the National Institute for Materials Science (NIMS) of Japan. In our previous work, to make the PoLyInfo data machine-readable and further machine-understandable, we built PoLyInfoRDF to store these data in the standard Resource Description Framework (RDF) format and then defined its schema in the Shape Expressions (ShEx) language. When designing the schema, it is important to modularize the schema such that the common components are reusable. This is the objective of this study and is essential for efficiently defining schemas of the descriptors and properties, which constitute the core of PoLyInfo, a large collection of experimentally measured polymer characteristics. As an example of modularization, descriptors of the source-based name and molecular formula both include a string value, hence their schemas may well share (‘inherit’) the schema for string values, which would be defined once and subsequently reused throughout the entire set of schemas. Actually we noticed a considerable amount of common portions among schemas of descriptors and properties, and clarified a ‘schema hierarchy’ to reflect the above ‘inheritance’ relationships, separately from the ontological ‘concept hierarchy’. We then investigated the extent to which the adapted strategy was able to successfully define the PoLyInfoRDF schema. Under this schema hierarchy, inheritance mechanisms in ShEx played a significant role in sharing common portions effectively in a well-organized manner. We expect future developments based on our approach to contribute to the standardization of scientific data representation in RDF by providing a library of reusable schemas. We have developed a new method for modularizing scientific data schemas and managing them hierarchically, and demonstrated it in PoLyInfo. This paves the way for data fusion in materials chemistry.PoLyInfo是日本国立材料科学研究所（National Institute for Materials Science，NIMS）的高分子数据库。在前期工作中，为使PoLyInfo数据具备机器可读性并进一步实现机器可理解性，我们构建了PoLyInfoRDF，以标准资源描述框架（Resource Description Framework，RDF）格式存储此类数据，并使用形状表达式（Shape Expressions，ShEx）语言定义了其模式。在模式设计过程中，对模式进行模块化以实现通用组件的复用至关重要，这既是本研究的核心目标，也是高效定义构成PoLyInfo核心的描述符与属性模式的必要前提——PoLyInfo作为收录大量实验测得的高分子特性的大型数据集，其描述符与属性正是该数据库的核心内容。以模块化实践为例，基于来源的名称描述符与分子式描述符均包含字符串值，因此二者的模式可共享（“继承”）字符串值模式，该模式仅需定义一次，即可在全部模式集内复用。事实上，我们发现各类描述符与属性的模式中存在大量共通部分，并厘清了一套“模式层级结构”以反映上述“继承”关系，且该层级结构独立于本体论层面的“概念层级”。随后我们评估了该适配策略在定义PoLyInfoRDF模式时的实际效果。在此模式层级结构下，ShEx中的继承机制可高效且有序地实现共通部分的共享，发挥了关键作用。我们期望基于本研究方法的后续工作，能够通过提供可复用模式库，推动RDF格式科学数据表示的标准化进程。本研究开发了一套可对科学数据模式进行模块化设计并层级化管理的新方法，并在PoLyInfo数据集上完成了验证。该方法为材料化学领域的数据融合奠定了基础。