摘要:This paper describes a computational implementation of the Finnish numeral system as
a single finite-state transducer that maps inflected numerals to the corresponding
numbers with tags that indicate morphological features, ordinality, number and case.
The transducer is bidirectional. It can be used to analyze complex numerals such as
kahdensienkymmenensienyhdeksänsien as 29+Ord+Pl+Gen and to generate from a
numeric input such as 251+Sg+Nom the corresponding inflected numeral kaksisataaviisikymmentäyksi.
The mapping from numbers to numerals in Finnish is much more
complex than a similar transduction for languages such as English because in Finnish
complex numerals are traditionally inflected and all parts agree in ordinality, number
and case, with the exception of nominative singulars such as kaksisataa ‘200’.
Nevertheless, a complete analyzer/generator for millions of Finnish numerals can be
built easily with the techniques described in the book on Finite State Morphology by
Beesley and Karttunen (2003) using the XFST and LEXC compilers.