Finite State Lazy Operations in NLP
Franck Guingne, Florent Nicart
Finite state networks can represent dictionaries and lexical relations. Traditional finite-state operations like
composition can produce huge networks with prohibitive computation space and time. For a subset of finite
state operations, these drawbacks can be avoided by using virtual networks, which rely on structures that
are partially buit on demand. This paper addresses the implementation of virtual network operations in xfst
(XEROX Finite State Technology software). The example of "priority union" which is particularly useful in NLP
, is developed.
Proc. of CIAA, Tours, France, July 3-5, 2002.