CTexT

ARTIST BIOGRAPHY

The Centre for Text Technology (CTexT™) at the North-West University (Potchefstroom Campus) is working with several partners towards the implementation of human language technology (HLT) in South Africa. HLT is a general term for information and communication technology that specialises in the interpretation and generation of speech and text in human language. Five CTexT employees are participating in the Artists Book project.

Ané Bekker has a MA-degree in art history and is the senior project manager. She manages projects that the Centre partakes in with partners such as the Department of Arts and Culture.

Cindy McKellar and Marissa van Rooyen both obtained their BA Language Technology degrees cum laude at the NWU, and are currently working as computational linguists, while completing MA-degrees in Language Technology.

Liza du Plooy recently completed her degree in BA Language Technology and plans to further her studies

Wildrich Fourie has been programming computers since age eleven and was recently invited to join a worldwide open source development project team.

Moira Müller is the communications officer and also obtained her honours degree in communication studies at the NWU.

Suléne Pilon manages the research activities at CTexT and teaches Afrikaans Linguistics and Language Technology at the NWU.

PROJECT

Title Lost in Translation
Medium Mixed media
Dimensions Varied
Edition Unique
Price -

Machine translation (MT) is a process during which a text in one natural/human language is automatically translated into text of another language (Hutchins, 1995).

MT was first conceptualised in 1947 by Warren Weaver and work on MT started in the early fifties. This was the beginning of computational linguistics, artificial intelligence, formal linguistics and non-numeric programming languages (Hutchins, 1995).

When MT proved to be a more complex task than was initially anticipated, almost all work in this field was abandoned after the publication of the ALPAC report (1966), which advised governments against investment in MT research. The 1980s saw a revival of work in MT and since then MT has come to be known as the holy grail of HLT, as it has been impossible to automatically generate a translation that is comparable to a human translation (Dirix et al., 2005).

The errors produced by MT systems vary in nature and severity, and include incorrect/inconsistent translations (i.e. distortion of the message) and poor target language documents (i.e. distortion/disregard of the linguistic features of a language).

This project explores reasons for the inadequate performance of MT by highlighting one of the main differences between man and machine, which is sensory perception, by focussing on aspects that are lost in automatic translation. Since a computer lacks senses, it is impossible for it to gain experience from the world around it and therefore also, from the viewpoint of experientialism, to acquire knowledge. Therefore, even though an MT system is based on various linguistic and programming models, it lacks the knowledge gained through sensory experience; knowledge which is necessary to translate correctly and accurately.

CTexT hypothesises that MT can be perfected by giving/teaching a computer sensory perception. This will enable it to gain the experience and knowledge needed to deliver output comparable to that of human translators.