The ODI’s Data Taxonomy

#TODO: Add “equity” and “fairness” considerations for each data item.

See the original document: https://theodi.org/news-and-events/blog/a-data-for-ai-taxonomy/

Category	Type of Data	Description
Developing AI systems	Existing data	Data not directly used for model training but as the basis for creating training datasets.
Developing AI systems	Training data	Data processed to train AI models by helping them recognize patterns and improve accuracy.
Developing AI systems	Reference data	Data used to enrich training datasets with context, such as knowledge graphs or linguistic resources.
Developing AI systems	Fine-tuning data	Smaller datasets used to adapt pre-trained models for specialized tasks while preserving their capabilities.
Developing AI systems	Testing and validation data	Data used to test models during development to ensure accuracy and representativeness.
Developing AI systems	Benchmarks	Datasets used to evaluate a model’s performance and accuracy against unseen data.
Developing AI systems	Synthetic data	Algorithmically generated data used for training, fine-tuning, or benchmarking models.
Developing AI systems	Data about the data	Information about the datasets used to develop AI models, such as their size, source, and composition.
Deploying AI systems	Model weights	Numerical values representing the relationships learned by a model during training.
Deploying AI systems	Local data	Data an AI model processes in a specific deployment context, depending on its purpose and architecture.
Deploying AI systems	Prompts	Instructions or queries given to AI systems to generate responses, commonly in generative models.
Deploying AI systems	Outputs from models	Generated data from AI systems, such as text, audio, video, or structured outputs.
Monitoring AI systems	Data about models	Information disclosed about AI models, including version, performance, and ethical considerations.
Monitoring AI systems	Data about model usage and performance in context	Data collected during model use, such as query logs and performance metrics, used for improvements.
Monitoring AI systems	Registers of model deployments	Authoritative lists of AI models deployed in specific contexts or sectors, maintained by governments or organizations.
Monitoring AI systems	Data about the AI ecosystem	Data about the broader AI ecosystem, including models, incidents, policies, and workforce statistics.