EBMT (Example-Based Machine Translation) is proposed. EBMT retrieves similar examples (pairs of source phrases, sentences, or texts and their translations) from a tahase of examples, adapting the examples to franslate a new input. EBMT has the following features: (1) It is easily upgraded simply by inputting appropriate examples to the database; (2) It assigns a reliability factor to the translation result; (3) It is accelerated effectively by both indexing axi parallel computing; (4) It is robust because of best-match reasoning; (5) It well utilizes translator expertise. A prototype system has been implemented to deal with a difficult Iranslation problem fee conventional Rule-Based Machine Translation (RBMT), i.e., translating Japanese noun phrases of the form 'lq a no N2" into English. The system has achieved about a 78% success rate on average. This paper explains the basic idea of EBMT, illustrates the experiment in detail, explains the broad applicability of EBMT to several difficult translation problems fee RBMT discusses the advantages of integrating EBMT with RBMT.

Ideally, we might hope to improve the performance of our MT systems by improving the system, but it might be even more important to improve performance by looking for a more appropriate application. A survey of the literature on evaluation of MT systems seems to suggest that the success of the evaluation often depends very strongly on the selection of an appropriate application. If the application is well-chosen, then it often becomes fairly clear how the system should be evaluated. Moreover, the evaluation is likely to make the system look good. Conversely, if the application is not clearly identified (or worse, if the application is poorly chosen), then it is often very difficult to find a satisfying evaluation paradigm. We begin our discussion with a brief review of some evaluation metrics that have been tried in the past and conclude that it is difficult to identify a satisfying evaluation paradigm that will make sense over all possible applications. It is probably wise to identify the application first, and then we will be in a much better position to address evaluation questions. The discussion will then turn to the main point, an essay on how to pick a good niche application for state-of-the-art (crummy) machine translation.- 2-1.

This paper is an attempt to survey the field in this period and to provide, in effect, an updating of the historical review completed in late 1984 and early 1985 (Hutchins 1986). The aim is to give a general picture of the recent developments of systems and projects already established at that time and to document the emergence of new systems and projects worldwide. Descriptions of individual systems are given in section II. This part outlines the main issues and lines of development at the present time. There is no claim of completeness, but it is hoped that all significant activities have been noticed

Abstract not found

It is now acknowledged that the evaluation of the quality of MT output is inextricably linked to the purpose to which the translation output will be put. It is also true to say that the value of the evaluation is inseparably linked to the purpose to which the evaluation results will be put. For the developer of an MT system, the evaluation of the quality of the MT output must be approached from the viewpoint of increas-ing the knowledge about the MT system. Resultant measures must be analysed, so that practical feedback to improve the system is feasi-ble. However, for the manager, evaluation of the MT output is often viewed in terms of comparison. Measures are compared against pre-vious measures, or against those obtained by other systems, in order to gauge progress, or to assess the systems ability. Measurement can be viewed as a tool for increasing the knowl-edge of some object or entity. From both viewpoints described above, the measurement is required as a means to increase the knowledge about the system, whether it is knowledge about the systems errors, or performance. However, there is a more fundamental level at which measurement should be applied as a tool for increasing knowledge; that is, to in-crease the knowledge of the properties we are trying to measure (in this case intelligibility and fidelity). Such measurement is a precursory requirement for more general uses of evaluation measures, as described above. When surveying the many methods currently employed in MT eval-uation, it is not immediately obvious that the methods used serve to increase the knowledge of the properties being measured. This re-port describes a constructive machine translation evaluation method, aimed at addressing this issue. 99

Abstract not found