Miscellaneous
Label
Labels are the "ground truth" outputs attached to training examples. In image classification, labels are class names; in NER, labels are entity type tags; in sentiment analysis, labels are positive/negative/neutral. Model training minimizes the gap between predicted and labeled outputs.
Label quality directly impacts model quality. Noisy labels (incorrect annotations) degrade learning; systematic labeling bias leads to biased models. Professional annotation services and inter-annotator agreement measurement are standard practices for high-stakes applications.
Authority Links
Related Terms
Miscellaneous
Training Data
The labeled or unlabeled dataset used to fit a model's parameters during the learning process.
Core Concepts
Supervised Learning
Models trained on labeled data, learning to predict outcomes from inputs.
Miscellaneous
Dataset
An organized collection of data examples prepared for training, evaluating, or testing AI models.
Techniques & Methods
Entity Annotation
Labeling text spans with entity type information to create structured training data.