This chapter introduces a concrete syntax. It is ideally isomorphic to an abstract syntax that specifies a semantic annotation scheme, while providing a format for representing annotation structures. This format can represent them either in a serialized way from left to right, or in graphic images or tabular forms with linking arrows. Representation formats may vary, depending on kinds of the use of annotation. Human readers, for instance, prefer tabular formats especially for illustrations or demonstrations. For the purposes of merging, comparing, or exchanging various types of annotations or different annotations of the same type, graphs are considered useful. In this chapter, I introduce a graphic annotation format, called GrAF, for linguistic annotation. For the construction of larger corpora, however, there are practical computing reasons to prefer a serialization of annotations. For the serialized representation of annotation structures, this chapter mainly discusses two formats: (i) XML and (ii) pFormat, a predicate-logic-like representation format, which represents annotation structures in a strictly serial (linear) manner by avoiding embedded structures.

In Chapter 9, I introduce eXTimeML, an extended variant of ISO-TimeML, with three extensions. (i) Temporal measure expressions are annotated as part of generalized measure: e.g., 30 hours. (ii) Quantified temporal expressions are annotated as part of generalized quantification: e.g., every day. (iii) Adjectives and adverbs are annotated as modifiers of nouns and verbs, respectively: e.g., daily, never. Temporal measures and temporal quantifiers are treated as part of generalized measures and quantifiers. I then illustrate how the representation language of ABS applies to each of these extensions in eXTimeML by deriving appropriate (logical) semantic forms from the well-formed annotation structures of temporal measures, quantifiers, and modifiers. Semantic forms are then interpreted with respect to admissible models, constrained by the formal definitions of logical predicates such as twice or three thousand.

In this chapter, I introduce the four types of category path: static, dynamic, oriented, and projected, while characterizing them for the interpretation of path-related information in language. Static paths are finite paths with two endpoints, but neither of the endpoints is identified intrinsically as the start or the end of a path. Dynamic paths are trajectories caused by motions. Oriented paths are simply directed to some goals and may not reach the goals. Projected paths are virtual or intended, which are not actually traversed but devised in the mind of a human or rational agent. To discuss their characteristic features in formal terms, I introduce Pustejovsky and Yocum’s (2013) axioms on motions and derive a corollary based on them. This corollary relates a mover to an event-path. I then show how the movement link (moveLink) is reformulated to link a mover to a motion-triggered event-path with the relation traverses. I also analyze the notions of orientation and projection with respect to the frames of reference, either absolute, relative, or intrinsic, while showing how these frames apply to the annotation and interpretation of oriented or projected paths.

This chapter works toward the specification of a dynamic annotation scheme, called dSpace. It extends the scope of ISO-Space to the domains of space and time over motions by being amalgamated with ISO-TimeML. In dSpace, various types of temporal relations interact with spatial relations. The temporal dimension characterizes various types of paths and motions anchored to each location on the paths; dSpace also generalizes the notion of paths by classifying them into four types: static, dynamic, projected, and oriented, while introducing a relational link, called pathLink, over paths with various relation types such as meet and deviate.

Recognition skills refer to the ability of a practitioner to rapidly size up a situation and know what actions to take. We describe approaches to training recognition skills through the lens of naturalistic decision-making. Specifically, we link the design of training to key theories and constructs, including the recognition-primed decision model, which describes expert decision-making; the data-frame model of sensemaking, which describes how people make sense of a situation and act; and macrocognition, which encompasses complex cognitive activities such as problem solving, coordination, and anticipation. This chapter also describes the components of recognition skills to be trained and defines scenario-based training.