TY - GEN

T1 - Building Intelligent Digitial Assistants for Speakers of a Lesser-Resourced Language

AU - Jones, Dewi

AU - Cooper, Sarah

N1 - Conference contribution does not have an ISSN, therefore not required to meet REF OA requirements.

PY - 2016/5/23

Y1 - 2016/5/23

N2 - This paper reports on the work to develop intelligent digital assistants for speakers of a lesser-resourced language, namely Welsh. Such assistants provided by commercial vendors such as Apple (Siri), Amazon (Alexa), Microsoft (Cortana) and Google (Google Now) are allowing users increasingly to speak in natural English with their devices and computers in order to complete tasks, obtain assistance and request information. We demonstrate how these systems’ architectures do not provide the means for external developers to build intelligent speech interfaces for additional languages, which, in the case of less resourced languages, are likely to remain unsupported.Consequently we document how such an obstacle has been tackled with open alternatives. The paper highlights how previous work on Welsh language speech recognition were improved, utilized and integrated into an existing open source intelligent digital assistant software project. The paper discusses how this work hopes to stimulate further developments and include Welsh and other lesser resourced languages in as many developments of intelligent digital assistants as possible.

AB - This paper reports on the work to develop intelligent digital assistants for speakers of a lesser-resourced language, namely Welsh. Such assistants provided by commercial vendors such as Apple (Siri), Amazon (Alexa), Microsoft (Cortana) and Google (Google Now) are allowing users increasingly to speak in natural English with their devices and computers in order to complete tasks, obtain assistance and request information. We demonstrate how these systems’ architectures do not provide the means for external developers to build intelligent speech interfaces for additional languages, which, in the case of less resourced languages, are likely to remain unsupported.Consequently we document how such an obstacle has been tackled with open alternatives. The paper highlights how previous work on Welsh language speech recognition were improved, utilized and integrated into an existing open source intelligent digital assistant software project. The paper discusses how this work hopes to stimulate further developments and include Welsh and other lesser resourced languages in as many developments of intelligent digital assistants as possible.

M3 - Conference contribution

SP - 74

EP - 79

BT - Proceedings of the LREC 2016 Workshop “CCURL 2016 – Towards an Alliance for Digital Language Diversity”

A2 - Soria, Claudia

A2 - Pretorius, Laurette

A2 - Declerck, Thierry

A2 - Mariani, Joseph

A2 - Scannell, Kevin

A2 - Wandl-Vogt, Eveline

ER -