Automatic Detection of Lexicalised Phrases in Swedish
Janne Lindberg · DSpace repository (University of Tartu) · 2000
I wiIrpresent a system under development, called LP-DETECT.The system detects and analyses Swedish lexicalised phrases (LPs) in order to enhance subsequent parsing.LPs are one of a number of stumbling blocks related to word sequences that must be dealt with when parsing unrestricted text.LPs include semantic idioms, syntactic idioms and morphological idioms and so called valency breaking LPs.The system reported on consists of an LP lexicon of some 8000 LPs with analyses, a detection program written in perl and rules for disambiguating between and discarding LP analyses.A small evaluation of the system is also presented.