TransType: a Computer--Aided Translation Typing System

This paper describes the embedding of a statistical translation system within a text editor to produce TRANSTYPE, a system that watches over the user as he or she types a translation and repeatedly suggests completions for the text already entered. This innovative Embedded Machine Translation system is thus a specialized means of helping produce high quality translations. 1 I n t r o d u c t i o n TRANSTYPE is a project set up to explore an appealing solution to the problem of using Interactive Machine Translation (IMT) as a tool for professional or other highly-skilled translators. IMT first appeared as part of Kay's MIND system (Kay, 1973), where the user's role was to help the computer analyze the source text by answering questions about word sense, ellipsis, phrasal attachments, etc. Most later work on IMT, eg (Blanchon, 1991; Brown and Nirenburg, 1990; Maruyama and Watanabe, 1990; Whitelock et al., 1986), has followed in this vein, concentrating on improving the question/answer process by having less questions, more friendly ones, etc. Despite progress in these endeavors, systems of this sort are generally unsuitable as tools for skilled trans][ators because the user serves only as an advisor, with the MT components keeping overall control over the translation process. TRANSTYPE originated from the conviction that a better approach to IMT for competent translators would be to shift the focus of interaction from the meaning of the source text to the form of the target text. This would relieve the translator of the burden of having to provide explicit analyses of the source text and allow him to translate naturally, assisted by the machine whenever possible. In this approach, a translation emerges from a series of alternating contributions by human and machine. The machine's contributions are basically proposals for parts of the target text, while the translator's can take many forms, including pieces of target text, corrections to a previous machine contribution, hints about the nature of the desired translation, etc. In all cases, the translator remains directly in control of the process: the machine must respect the constraints implicit in his contributions, and he or she is free to accept, modify, or completely ignore its proposals. So TRANSTYPE is a specialized text editor with an embedded Machine translation engine as one of its components. In this project we had to address the following problems: how to interact with the user and how to find appropriate multi-word units for suggestions that can be computed in real time. 2 T h e T r a n s T y p e m o d e l 2.1 U s e r V i e w p o i n t Our interactive translation system is illustrated in figure 1 for an English to French translation. It works as follows: a translator selects a sentence and beg!ns typing its translation. After each character typed by the translator, the system displays a proposed completion, which may either be accepted using a special key or rejected by continuing to type. This interface is simple and its performance may be measured by the proportion of characters or keystrokes saved in typing a translation. Note that, throughout this process, the translator remains in control, and the machine must continually adapt its suggestions to the translator's input. This differs from the usual machine translation set-ups where it is the machine that produces the first draft which