Training Data

The collection of labeled examples used to teach a machine learning model to perform a specific task. The quality, quantity, and diversity of training data directly determine how well a model will perform. In accessibility contexts, training data quality is especially important because biases or gaps in the data can lead to AI systems that fail for users with disabilities — for example, speech recognition trained primarily on non-disabled speakers may struggle with atypical speech patterns. When blind users create their own training data for teachable object recognizers, ensuring photo quality without visual verification requires alternative feedback mechanisms like data descriptors.