A Syntactic Time-Series Model for Parsing Fluent and Disfluent Speech