Interpreting motion: Grounded representations for spatial language by Inderjeet Mani, James Pustejovsky (review)

The treatment of space and spatial language has long been an area of philosophical, theoretical, and now also increasingly practical concern. Detailed studies of spatial language and its use have been pursued in the areas of language typology, formal semantics, psycholinguistics, child language development, language relativity, human-machine interaction, dialogue analysis, computational modeling, geographical information science, formal ontology, and many more. Interaction with others in, and about, space is a fundamental aspect of human intelligent behavior, and explicating the role of language in this capacity raises substantial challenges. Inderjeet Mani and James Pustejovsky’s volume for the Oxford University Press series ‘Explorations in language and space’ offers a contribution to one component of this broad area of investigation by developing a treatment of spatial language expressions involving ‘motion’.

The essential idea of M&P’s approach is straightforward. A fairly standard syntactic analysis is used for the compositional construction of a denotational event-based semantics involving spatial categories. The lexical argument structure of verbs and other syntactic categories involving motion introduces semantic categories such as ‘path’, ‘location’, ‘orientation’, ‘figure’ (an object to be located, or ‘locatum’), and ‘ground’ (a background or reference object against which or relative to the locating occurs, or ‘relatum’). Further conditions involving the manner of movement, the medium through which the movement occurs, and the path over which the movement occurs are then specified as changes in the spatial relationships holding over the entities identified. A verb such as bounce, for example, is characterized as an activity where the figure is first in an externally connected spatial relationship to some ground and subsequently in a spatial relationship of ‘disconnection’ to that ground repeatedly throughout the activity; similarly, the verb leave receives a semantics indicating that the figure is first ‘in’ some identified ground (e.g. a ‘room’) and at a later time is spatially disconnected with that ground.

The structure of the book and the authors’ presentation of their material follow broadly their respective interests. P has been concerned with articulating formal semantic descriptions for selected areas of natural language for many years, while M has addressed the tasks and challenges of providing practical annotation schemes and corresponding tools for large-scale corpora. Previously, these interests were combined in a corpus markup scheme for temporal expressions, called TimeML, that took proposals for the semantics of natural language temporal expressions as the basis for semantically well-defined annotations. A move to include spatial language in the same spirit then appeared promising, and so the methodologies and tools developed for time were extended, first to the annotation of static spatial expressions (SpatialML) and, with this book, to motion expressions (now also presented as part of a proposed markup standard for spatial annotation of corpora called ISO-Space).

M&P build on two theoretical pillars. First, they import results from the area of qualitative spatial representation to provide the spatial relations they need and their formalizations (Ch. 3), and second, the resulting semantic expressions are couched in terms of a dynamic interval logic that allows for the definition of procedures of the kind just suggested for bounce and leave (Ch. 4). M&P then suggest that the representations developed are a sound basis for large-scale corpus annotation of motion events (Ch. 5) and offer some illustrative applications showing the value of both the formal framework and its provision of richly annotated linguistic data (Ch. 6).

This development draws primarily on work in two areas. On the one hand, there is substantial linguistic research from researchers such as Ray Jackendoff, Leonard Talmy, Ronald Langacker, Annette Herskovits, and many more addressing the linguistic expression of space, proposing linguistic semantics of space in a variety of frameworks, addressing consequences of compositionality, and so on. On the other hand, there is also extensive formal work exploring the qualitative representation of space by means of well-specified ‘calculi’ that allow specifications of the relationships...