语言资源管理 语料库标注项目管理 第1部分：核心模型

[中文适用范围]： 本文件建立了语料库标注项目管理的核心模型，规定了项目团队的工作包、所需流程和可交付成果。本文件针对协调、人员培训、可重用性、软件、质量控制、许可和版权等问题提供了必要的组成部分。然而，它并未规定解决此类问题的方法论。本文件就项目所需的工作包和可交付成果提供指导，这些工作流程和流程处理以下内容： — 工作包之间的集成与沟通：包括确保所有工作包良好协调，特别是在采用更广泛的标注标准和与本体集成以增强互操作性方面。跨工作包的有效沟通对于与其他项目无缝共享标注文档至关重要。 — 人力资源管理和评分者间可靠性：涵盖人力资源管理，侧重于培训和资格认证，以及实施评分者间可靠性实践。这些实践包括培训、测试和使用适当的工具以确保标注的一致性。 — 标注指南管理和软件利用：涉及管理标注任务的指南，并利用标注软件和工具，特别是在利用人工智能和机器学习技术的环境中。 — 质量控制、数据验证和结构化文档：包括对标注结果进行质量控制和验证的流程，以及结构化文档和持续管理的需求。这确保了标注文档长期保持准确、相关和可用。 — 许可、版权和元数据管理：侧重于记录许可和版权，提供元数据以管理资源的共享。在存在版权限制或许可问题的领域尤为重要，确保数据子集可以得到适当管理和共享。 [外文原描述]： This document establishes a core model of project management for corpus annotation, to specify the work packages of project teams, required processes and deliverables. This document presents the necessary components for issues such as coordination, human training, reusability, software, quality control, licensing and copyright. However, it does not specify a methodology to solve such issues. This document gives guidance on what work packages and deliverables are required under the project in which workflows and processes deal with the following: — Integration and communication among work packages: This includes ensuring that all work packages are well-coordinated, particularly in terms of the adoption of broader annotation standards and integration with ontologies to enhance interoperability. Effective communication across work packages is crucial for the seamless sharing of annotated documents with other projects. — Human resource management and interrater reliability: This covers the management of human resources, focusing on training and qualification, as well as the implementation of interrater reliability practices. These practices include training, testing and the use of appropriate tools to ensure consistency across annotations. — Annotation guideline management and software utilization: This involves managing the guidelines for annotation tasks and utilizing annotation software and tools, particularly in environments leveraging artificial intelligence (AI) and machine learning (ML) techniques. — Quality control, data validation and structured documentation: This encompasses the processes for quality control and validation of annotation results, alongside the need for structured documentation and ongoing curation. This ensures that annotated documents remain accurate, relevant and usable over the long term. — Licensing, copyrights and metadata management: This focuses on documenting licences and copyrights, providing metadata to manage the sharing of resources. It is particularly important in areas with copyright restrictions or licensing concerns, ensuring that data subsets can be appropriately managed and shared.