Natural language processing for under-resourced languages: Developing a Welsh natural language toolkit

Language technology is becoming increasingly important across a variety of application domains which have become common place in large, well-resourced languages. However, there danger that small, under-resourced languages are being pushed to the technological margins. Under-resourced face significant challenges delivering underlying language resources necessary support such applications. This paper describes development natural processing toolkit for an language, Cymraeg (Welsh). Rather than creating Welsh Natural Toolkit (WNLT) from scratch, approach involved adapting and enhancing functionality provided other within existing framework making use external where available. begins by introducing GATE NLP framework, was used as platform WNLT. It then each core modules WNLT turn, detailing extensions adaptations required processing. An evaluation reported. Following this, two demonstration applications presented. The first simple text mining analyses wedding announcements. second Twitter application, extends pipeline. As relatively small-scale project, makes possible, rather new resources. adaptation reuse can provide practical achievable route developing