A novel way of extracting full-fledged terminologies from multilingual texts

Abstract

Abstract Terminology as a term refers to terminology science and simultaneously to practically curated terminological resources (TR). The latter fall within the scope of language resources. Manually curating such a TR is labor- and time-intensive. Thus, it has been a long-standing research endeavor to reduce this effort by automating the TR creation process. However, to this point such endeavors have focused on terms rather than on interrelations. Instead of directly interrelating terms, they are conventionally grouped by synonymy and equivalence into concepts, and relations are established between concepts. This paper introduces preliminaries of terminology science and practice, common models of interrelating terminological concepts, and especially focuses on a new way to define terminological relations and a system to extract terminological information from texts across domains and languages by means of recent neural methods.