Purpose: domains and modalities describe the information and problems that models work with. A domain is a field of tasks, such as Computer Vision (CV) or Natural Language Processing (NLP); a modality is a form of data, such as images, text, audio, or video.
Examples: computer vision analyzes visual information, NLP processes human language, speech and audio systems analyze or generate sound, and multimodal models connect multiple data forms. These areas overlap: speech recognition connects audio and language, while image captioning connects vision and language.