ABSTRACT
This chapter presents the essential requirements for natural language understanding ability of cognitive robots. A natural language understanding system is destined to respond in a certain style or protocol in order to show the evidence of its valid understanding of the input natural language expression. Conventional approaches to natural language understanding for robots depend on certain special knowledge of the task domains because the implemented semantics are too naive to concern robotic sensation and action. Dependency structures are well suited for the analysis of languages with free word order, such as Japanese, Czech, and Slovak. The natural language understanding system is intended to detect ill-formed inputs at every level of processing, namely, ungrammatical at syntactic analysis, anomalous at semantic analysis, or impossible at pragmatic analysis. Well-elaborated text or speech corpora specific to the discourse domains are helpful for a robust natural language understanding system to retrieve candidates for correction based on morphological or phonological similarity in advance to semantic evaluation.
