Towards a Ubiquitous Language Resource Management System
Mohohlo Samuel Tšoeu, S. Maxaku, T. Chemvura, N. Ramchunder, Z. Mabuza · 2019
Natural Language Processing has a strong need for large scale quality speech, sign and text resources. The problem in developing countries in African and globally, is the lack of such resources. Significant progress has been made by research groups in South Africa under the coordination Language Resource Management Agency, to collect language resources for the 11 official languages of South Africa. Existing limitations are that resources have been collected manually, in a unidirectional manner that only benefits research institutions and not the wider community that needs them the most and the data is statically pre-packaged and not dynamic. In this paper we detail the design and initial implementations of a ubiquitous language resource management system. The system is aimed at supporting manual collection, crowdsourcing and distribution of language resources and to provide a dynamic resource management system that allows submission, acquisition and quality auditing of languages resources based on user specifications as opposed to pre-packaged data formats. We present initial implementations of the server side and mobile application clients and outline future developments and system integration.