Deciphering the Nomenclature of Chemical Compounds in NooJ
摘要
This paper proposes a multifaceted approach to the design of an algorithm for the automatic recognition of chemical compounds in Croatian written as multiword expressions. The algorithm, which we have named the Croatian Chemical Compounds Module, consists of three layers: it uses (1) the NooJ dictionary as the basis for (2) a morphological grammar, and both (1) and (2) are used for (3) a syntactic grammar. This module supports not only single-unit words and homoatomic entities but also variations of chemical names recognized through a variety of suffixes, multiplicative prefixes, hyphens, Roman and Latin numerals, Greek letters, and round, square and curly brackets. Terminological diversity and inconsistency in writing style are discussed as they present a great problem for any such endeavor.