Research seminar on computational linguistics

A multi-, inter- and trans-disciplinary (MIT) research seminar on Computational linguistics for lesser-resourced languages was hosted by the Department of African Languages in collaboration with the School of Computing on 17 July 2009. Dr Kenneth R Beesley from SAP Labs, Salt Lake City, USA, a computational linguist who has been working on finite-state natural-language processing for over 20 years, and Mr Jackson Muhirwe, a PhD student in Computer Science from the Makerere University in Uganda, presented tutorials to an audience of academics and postgraduate students principally from the Departments of African Languages, Linguistics and the School of Computing Sciences. The seminar was also attended by interested persons from the Meraka Institute’s Human Language Technologies and from a non-profit organisation focused on the localisation of open source software into South Africa's 11 official languages.

Dr Beesley gave an overview of the field of Computational Linguistics in Where computers and natural languages meet, arguing that there are many interesting applications, both academic and commercial, requiring interdisciplinary cooperation from formal linguists, language experts and computer scientists. He also suggested four textbooks that could be used in courses to help linguists become better acquainted with computational techniques, and to help computer scientists to become more aware of linguistics.

Mr Muhirwe demonstrated a practical application of computational linguistics by elaborating on his PhD research about Finite state morphology for Kinyarwanda, a language that is closely related to the African languages spoken in South Africa.

The seminar was supplemented by curriculum-planning discussions of the Department of African Languages together with colleagues from the School of Computing on 20 July 2009, in which Dr Beesley and Mr Muhirwe participated actively. These discussions are bound to give content and effect to the vision of Unisa’s role as an African university with an educational outreach and scholarly networks across the continent.

On 21 July 2009, Dr Beesley gave an invited talk to the special session on African and under-resourced languages at the Finite State Methods and Natural Language Processing (FSMNLP) conference, held this year in Pretoria (http://fsmnlp2009.fastar.org/Home.html).

He argued that finite-state techniques, implemented in a variety of accessible software packages, could be and have been productively used to further the documentation and promotion of under-resourced and generally neglected languages, definitely including the African languages of South Africa.

This was Dr Beesley’s fourth visit to South Africa and Unisa, each time at the invitation of Professors Sonja Bosch (Chair of Department of African Languages) and Laurette Pretorius (School of Computing), who developed an advanced finite-state morphological analyser for Zulu, and who have extended their work, through education and "bootstrapping", to most of the other official languages in South Africa. According to Dr Beesley: “They have become skilled and innovative developers, expert advocates of finite-state techniques, valued colleagues...” They, and Mr Muhirwe, who is working on the Kinyarwanda language, use the Xerox/PARC Finite State Toolkit, which Dr Beesley helped to develop and document in his co-authored book Finite State Morphology (CSLI Publications, 2003).

The visit of Dr Beesley and Mr Muhirwe was made possible with the generous financial support of the College of Human Sciences and assistance from the research fund of the School of Computing.