People rarely speak using the full rules of English grammar: that, however, does not stop us from understanding speech. This points to the fact that everyone uses some common constructs during spontaneous speech and that these constructs form the basis of everyone's understanding of spoken English. An in-depth analysis of a large corpus of spontaneous spoken English, in particular the spoken lecture, has been carried out which concentrated on identifying these common constructs and indicating problems posed by speakers of naturally spoken English. The results of the analysis show that these common constructs do exist. These islands of grammar, along with the features needed to deal with the unique problems posed by spontaneous speech, can be linked to form the major part of a speech grammar. This paper describes the analysis carried out and presents the results of this analysis. The overall aim of this work is to include these findings into a speech recognition system being developed at the University of Durham.
Keywords: continuous speech recognition, spontaneous speech, speech grammar
Bibliographic reference. Garigliano, Roberto / Johnson, Kevin / Collingham, Russell J. (1993): "A data-driven case for a spontaneous speech grammar", In EUROSPEECH'93, 969-972.