Welsh Natural Language Knowledge Base

Elliot Howells

Supervised by Alun D Preece, Moderated by Michael Daley

CENode (Controlled English Node - http://cenode.io) is a lightweight natural language knowledge-based system implemented using JavaScript, designed to run effectively in a wide variety of contexts, from servers to mobile devices and Internet ???things???. CENode-based apps interact with users in natural language using a conversational protocol that allows people to input queries and pieces of information (ask and tell); the apps can also ask and tell the user things. Example CENode-based apps include Moira (Mobile Information Reporting Agent) for obtaining ???on the spot??? reports (e.g. from people on patrol) and SHERLOCK (Simple Human Experiment Regarding Locally Observed Collective Knowledge) for crowdsourcing knowledge bases. The CENode knowledge base uses a form of restricted English, called Controlled English (developed by IBM). The goal of this project is to create an alternative version of this language, based on Welsh, and to adapt CENode to process ???Controlled Welsh??? (no advanced knowledge of JavaScript will be needed for this). The project will identify one or more applications for ???Controlled Welsh??? and create demonstration apps using the extended CENode software to demonstrate the usefulness of ???Controlled Welsh???, e.g. for language learners, for government apps, or for crowdsourcing information on current usage of the Welsh language.

Initial Plan

Final Report

Publication Form