ABSTRACT
This chapter clarifies the requirements for robotic natural language understanding viewed from systematic computation and presents concrete algorithms to satisfy them, focusing on 4D language. Spatiotemporal (i.e., 4D) language is the important tool for intuitive human-robot interaction. In such a situation, robots must understand natural language semantically and pragmatically well enough for comprehensible communication with humans. Knowledge representation languages (KRLs) for natural language understanding, including any semantic representation schemes, are seldom explicitly given the semantics of themselves but often implicitly, by employing natural language words in the vocabularies. The formal interpretation of a KRL is based on the set theory and the intuitive interpretation, exclusively for human users, is given in meta-use of a subset of natural language. Logical adequacy of mental image description language is mainly due to its expressive power, which in turn depends upon the set of constants specific to the mental image model because logical ones are granted for rigid and common to any logic-based KRL.
