Towards convergence of methods for speech and sign segmentation

Abstract

Signed languages, like spoken languages, combine sequences of arbitrary, finite, distinctive units in a continuous stream of movement (Stokoe 1960). Also like speech, determining where a particular sign begins and where it ends in a signing stream is not an easy task. Some researchers have established as the beginning of a sign the moment the hand is placed at the location where a certain sign is going to be initially or entirely produced and as its end the moment when the hand starts moving back to the rest position or to the location of the following sign (Crasborn & Zwitserlood 2008, Johnston 2009, Johnson & Liddell 2011). By doing so, these researchers leave transitional movements out of the limits of a sign. An alternative view claims that transitions should be partially or entirely regarded as part of a sign. Supporters base this view on the observations that (1) some articulatory features of a sign are visible even before or still after a sign is produced and (2) perceivers are able to guess signs solely drawing on information conveyed during transitions (Kita et al 2006, Bressem 2011, Jantunen 2010, 2013, 2015). The present study uses video data of Brazilian Sign Language (Libras; Xavier 2014) to critically evaluate the criteria traditionally used to delimit lexical items in the sign stream. Results indicate that methods used by speech researchers to delimit units in the speech stream are likely to be a good fit for delimiting units in the sign stream as well. Implications for speech and signed motor control will be discussed.