What is data augmentation, and how does it differ across modalities (images, text, audio)?

Data augmentation cheaply expands training data and regularizes models, but valid transforms differ by modality. The signal is the label-preserving constraint and why text is the hard one. Here is the answer.

Updated Aug 2026 · Grounded in real Applied AI Engineer interview loops and written to a senior-engineer editorial bar.