Michael Paul, Yasuharu Den
Stochastical Parsing of
Ill-Formedness in
Spoken Natural Language
Abstract:This report describes a statistical approach to parse spontaneous speech, taking into
account the phenomena of natural conversations. The input sentences are taken from
the ATR Dialogue Database. Using an adapted version of the stochastical language
model BLI, we yields in the detection and recovery of ill-formedness occuring in
spontaneous speech.