Development of automatic detection method for speech disfluencies in spontaneous Russian speech

Annotation: An algorithm for automatic detection of most frequent speech disfluencies and artefacts in the spontaneous Russian speech was proposed basing on the analysis of acoustical features of elements in question. The method of cross-correlation was used to detect voiced hesitation phenomena and a method of band-filtering was used to detect unvoiced hesitation phenomena and artefacts. The experiments were performed on especially collected corpus of 18 spontaneous Russian map-task and appointment-task dialogs. Experiments showed that voiced hesitation phenomena are detected with 80% accuracy and devoiced hesitation phenomena and artefacts – with 66% accuracy.

Keywords: speech disfluencies, speech corpus, automatic speech processing, automatic speech recognition