T/SAIAS 015-2024 Corpus Construction Guidelines

Scope

Main technical content: This standard mainly covers the relevant content of artificial intelligence corpus construction. It stipulates its general requirements and methods, and is applicable to a variety of related work. In terms of data resources, it requires large-scale, broad-coverage, and diverse data, with compliant confidentiality settings and high quality. For different data representation modes, such as text, image, audio, and video data, there are detailed regulations on categories and languages. The provision methods include physical storage media, and the data files are identified with naming rules. For corpus data, there are standardized requirements in general. The structure and content are divided into multiple sections with specific explanations, and the quality should meet standards. It is recommended to deliver the data using physical storage media. In the corpus production system, the overall architecture includes multiple layers. The corpus management system should have multiple functions, and the annotation tools of the toolchain need to meet minimum functional requirements. The corpus production process covers resource data inspection and the production process of text, image, audio, and video corpora. The value alignment of the corpus must be legal and compliant, with correct value orientation and continuous implementation. In terms of data security, the provider needs to implement various security strategies and measures, conduct security evaluations throughout the entire process, establish relevant systems, configure hardware and software components, and require participating personnel to sign confidentiality agreements and establish security teams.

Development Information

Related Standards