An MT learning environment for computational linguistics students

An MT Learning Environment for Computational Linguistics Students Svetlana Sheremetyeva Department of Computational Linguistics Copenhagen Business School, Bernhard Bangs Alle 17 B, DK-2000, Denmark lanaconsult@mail.dk Abstract This paper discusses the issue of suitability of software used for the teaching of Machine Translation. It considers requirements to such software, and describes a set of tools that have initially been created as developer environment of an APTrans MT system but can easily be included in the learning environment for MT training. The tools are user-friendly and feature modularity and reusability. Introduction A current trend in the teaching of Machine Translation to bridge the gap between theory and practice to a large extent consists of using on-line MT demos and/or commercial systems or other tools developed in the course of MT research. No matter that it has become possible only recently when in addition to the text books, computers and MT software have become an inherent part of MT learning environment; quite an experience has already been accumulated in developing an MT training task pool. Example exercises based on existing MT software can be found in practically every paper on the teaching of MT. See, for example, recent articles such as Pérez-Ortiz and Forcada (2001), Somers (2001), Clavier and Poudat 2001), Kenny and Way (2001), Blanc (2001). Still it is not that easy nowadays to provide for students to have a hands-on MT experience. The survey on Tools and Techniques for Machine Translation Teaching (Balkan et al., 1997) concludes that using a working MT system in the classroom involves a huge amount of effort both in terms of getting funds for purchasing commercial MT systems and in terms of using them to the most advantage of MT students which is not always possible due to the "locked" character of commercial products. This paper attempts to contribute to the problem by raising the issue of suitability of MT software for an MT student (and instructor). In other words, we believe that the MT learning environment will benefit from a kind of training software that was specially selected or designed for an MT student as the main user. From this perspective MT training software should be treated as any computer application for special purposes and thus should meet quite a number of requirements. A massive collection of papers on human- computer interaction - see, for example, Dix et al. (1998), Beyer and Holtzblatt (1998), Hackos and Redish (1998) - attempt to bring some structure to the often chaotic application design process. Topics include human limitations, usability principles, interface design, models of the user, task analysis, and practical methods of gathering data about users and tasks. It is stressed that there are no minor issues in the field of interface design and even such common topics as error messages, toolbars, tabbed dialogues, icons and responsiveness should be well thought out. In what follows we first attempt task and user analysis for an MT training tool as a special computer application for a Computational Linguistics student. We …

Content truncated to 500 words