A corpus-based approach to generalising a chatbot system

Abstract

International research in NLP is dominated by work on English. NLP techniques and systems can be ported to other natural languages, but this is generally a labour-intensive task, requiring scarce computational and linguistic expertise; hence minority languages are poorly represented in NLP technology. We present an automated approach to porting an NLP technology, the AIML-based chatbot, to new languages, by using a corpus in the target language to retrain the chatbot. We have successfully automated production of chatbots talking French, and Afrikaans; and are developing further demonstrators in Spanish and Arabic.