A bridge from the use-mention distinction to natural language processing

Abstract

Within computer science, the study of the syntax and semantics of metalanguage is well developed for formal languages, and this work is applied prominently in the creation of programming languages and compilers. However, relatively little work has been done in computer science to address metalanguage in natural languages. This lack has been to the detriment of language technologies that could exploit the information expressed in metalanguage to understand users’ utterances. This chapter addresses metalanguage and quotation from the perspective of mentioned language, a closely related phenomenon, and describes its relevance to core and applied work in natural language processing (NLP), a field in computer science concerned with the interaction between computers and natural languages. Examples are given for how state-of-the-art language technologies fail to cope with mentioned language. Finally, to promote progress on the computational study of mentioned language, a rubric is given for identifying the phenomenon in text. This enables human annotators to work methodically on labeling text to train NLP systems, a crucial precursor to further computational work.