Eyra - Speech Data Acquisition System for Many Languages

Abstract

Speech data acquisition is particularly important for under-resourced languages. The data gathering is the most labour-intensive part of developing speech technologies such as automatic speech recognizers and synthesizers. It is therefore important to facilitate this process with as much automation and labour-cutting tools as possible. This paper describes a new open-source system called Eyra which enables distributed speech data collecting through a variety of devices. It addresses internet connectivity issues by allowing the data collectors to run the back-end server off a local laptop, thereby facilitating automatic quality control and less labour-intensive data uploading and compiling. It can also be used in a crowd-sourcing set-up where volunteers can donate voice samples through a desktop web-browser interface. An initial test shows that the system works well in an offline mode using smart-phones for data collection.