Abstract
The notion of "perspective" is employed in language to introduce a shift in the modal accessibility to a situation by an agent (speaker/hearer). When annotating different kinds of linguistically related phenomena, the specification language (markup scheme) usually reflects the needs of the immediate task. For example, in a spatial annotation task, such as traversing a landscape or following a path, distinct frames of reference must be distinguished in order to correctly interpret the language of viewpoint: e.g., to your left, north of the castle, in front of the pub. For personal or ideological viewpoint annotation, on the other hand, the scheme must reflect the ability to position a "view" relative to different agents. Similar remarks hold for temporal perspective annotation, and conceptual perspectives. In this talk, we explore a way of representing point of view over any domain, using a modal logic of perceptual knowledge. Among other consequences, relative frame of reference expressions are interpreted as the composition of relational statements about two objects relative to the agent. We show how this scheme can be used to annotate point of view in diverse domains