The advancement of smart shipping and autonomous navigation relies on Automatic Identification System (AIS) data, which provides essential ship trajectory information. However, raw AIS data lacks semantic context, making behavior annotation crucial for understanding navigation tasks and processes. Existing research faces two challenges: (1) a lack of clarity on which semantics should be abstracted for effective behavior annotation, and (2) insufficient consideration of spatial interactions between ship maneuvering and the navigational environment, particularly topological interactions. These issues complicate data extraction and hinder machine learning-based applications such as explainable trajectory prediction. This paper proposes a comprehensive framework for semantic annotation and indexing of ship behavior. The framework deconstructs ship behavior into a unified data structure using a relational database, where three types of behavior semantics are defined, including atomic, topological, and traffic behavior. Atomic behaviors (e.g., move and stop) are extracted to annotate raw trajectories, while topological behaviors, describing interactions between trajectories and the environment, are modelled using an improved Dimensionally Extended 9-Intersection Model (DE-9IM). The combination of these semantics enables the annotation of higher-level traffic behavior. The model is further evaluated via behavior annotation statistics, demonstrating its effectiveness in annotation and indexing high-level ship behavior.