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.
