Annotation

Annotation is the addition of notes, comments, or metadata to a text, data object, image, or other content to explain, interpret, or categorize it. Annotations can be created by humans or produced automatically by software, and they may appear as inline notes, marginalia, or structured metadata.

In linguistics and the humanities, annotations clarify meaning, provide glosses, or mark features for analysis (for example, part-of-speech tags or named entities). In computing, annotations appear as code comments or language-specific annotations that influence behavior or provide metadata for tools. In data science and computer vision, annotations label samples (such as image labels or bounding boxes) to train machine learning models.

Annotation processes involve defining a scheme or schema, selecting tools, and ensuring consistency. Methods include manual annotation by experts, crowdsourcing, and automatic pre-annotation with human validation. Inter-annotator agreement is a common quality metric used to assess reliability.

Applications span digital humanities projects, data labeling for natural language processing and computer vision, software development, and research synthesis. Annotations support search, retrieval, interpretation, and reproducibility by documenting decisions and enabling structured access to information.

Challenges include subjectivity and ambiguity, scalability, and maintaining annotation quality over time. Best practices emphasize clear guidelines, version control, defined schemas, and transparent reporting of the annotation process and criteria.