Utilizing Evaluation in Networked Machine Translation*

In this paper we develop the prerequisites of a processing strategy to be employed in a networked machine translation environment. Networked Machine Translation (NMT) is a new MT paradigm based on the recent and emerging developments in the field of information technology (IT), World Wide Web (WWW) technology and language technology (LT), in particular networked computing. This new MT paradigm aims at combining the different technologies in an integrated way to achieve a better translation throughput in terms of translation quality and translation processing speed. The basic methodology of NMT is the deployment of cooperating and communicating software agents which make use of local and distributed information resources (different types of MT systems, language resources, user parameter, evaluation and verification parameters, and so forth) that are based on agreed standards (especially information exchange formats). The language resources of this new kind of MT software are intended to be shareable with other network-based language technology applications and services such as multilingual search engines, text retrieval, information brokering and telematics services. NMT can be deployed in intranet applications (corporate networks based on Internet and World Wide Web technologies) and is thus open to general translation services to the public users through the Internet.