Scarcity of digitally available language resources restricts development of large scale speech applications in Indian scenario. This paper describes a unique design framework for telephony speech data collection in under-resourced languages using interactive voice response (IVR) technology. IVR systems provide a fast, reliable, automated and relatively low cost medium for simultaneous multilingual audio resource collection from remote users and help in structured storage of resources for further usage. The framework needs IVR hardware & API, related software tools and text resources as its necessary components. Detailed functional design and development process of such a running IVR system are stepwise elaborated. Sample IVR call-flow design templates and offline audio transcription procedure is also presented for ease of understanding. Entire methodology is language independent and is adaptable to similar tasks in other languages and especially beneficial to accelerate resource creation process in under-resourced languages, minimizing manual efforts of data collection and transcription.
Cite as: Basu, J., Khan, S., Samirakshma Bepari, M., Roy, R., Pal, M., Nandi, S. (2018) Designing an IVR Based Framework for Telephony Speech Data Collection and Transcription in Under-Resourced Languages. Proc. 6th Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU 2018), 47-51, doi: 10.21437/SLTU.2018-10
@inproceedings{basu18b_sltu,
author={Joyanta Basu and Soma Khan and Milton {Samirakshma Bepari} and Rajib Roy and Madhab Pal and Sushmita Nandi},
title={{Designing an IVR Based Framework for Telephony Speech Data Collection and Transcription in Under-Resourced Languages}},
year=2018,
booktitle={Proc. 6th Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU 2018)},
pages={47--51},
doi={10.21437/SLTU.2018-10}
}