Metadata is a broad concept that describes data that explains or identifies other data. Metadata can be used for many purposes, from driving automation of document builds to enabling authors and readers to find content more easily. DITA provides a number of different types of metadata, each of which has different uses and is created in different places and ways. Some of the most important forms of metadata in DITA are topic and taxonomy.
Topic metadata describes the topic and what it is about. Topic metadata can be inserted in the prolog element of a topic or inside the topicref element that points to a topic from a map. In other words, metadata about the topic can be asserted by the topic itself, or can be assigned to it by the map that includes it in the build. This allows various different maps to assign metadata to the same topic. This may be appropriate where you want to describe a topic differently in various documents.
A taxonomy is a controlled vocabulary. It can be used to standardize how many things in your content and metadata are named. This consistency in naming can help ensure that automated processes work correctly, and that consistent terminology is used in content, and in metadata. In DITA, taxonomies are created using subject scheme maps. When you are authoring, many of the values you choose from have been defined in subject scheme maps.