TransAhead: A Computer-Assisted Translation and Writing Tool

2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 352–356, Montr´eal, Canada, June 3-8, 2012. c⃝2012 Association for Computational Linguistics TransAhead: A Computer-Assisted Translation and Writing Tool *Chung-chi Huang +Ping-che Yang **Keh-jiann Chen ++Jason S. Chang *ISA, NTHU, HsinChu, Taiwan, R.O.C. **IIS, Academia Sinica, Taipei, Taiwan, R.O.C. +III, Taipei, Taiwan, R.O.C. ++CS, NTHU, HsinChu, Taiwan, R.O.C. {*u901571,+maciaclark,++jason.jschang}@gmail.com; **kchen@iis.sinica.edu.tw Abstract We introduce a method for learning to predict text completion given a source text and partial translation. In our approach, predictions are offered aimed at alleviating users’ burden on lexical and grammar choices, and improving productivity. The method involves learning syntax-based phraseology and translation equivalents. At run-time, the source and its translation prefix are sliced into ngrams to generate and rank completion candidates, which are then displayed to users. We present a prototype writing assistant, TransAhead, that applies the method to computer-assisted translation and language learning. The preliminary results show that the method has great potentials in CAT and CALL with significant improvement in translation quality across users. 1 Introduction More and more language workers and learners use the MT systems on the Web for information gathering and language learning. However, web translation systems typically offer top-1 translations (which are usually far from perfect) and hardly interact with the user. Text translation could be achieved more interactively and effectively if a system considered translation as a collaborative between the machine generating suggestions and the user accepting or overriding on those suggestions, with the system adapting to the user’s action. Consider the source sentence “我們在完成這筆交 易上扮演重要角色” (We play an important role in closing this deal). The best man-machine interaction is probably not the one used by typical existing MT systems. A good working environment might be a translation assistant that offers suggestions and gives the user direct control over the target text. We present a system, TransAhead1, that learns to predict and suggest lexical translations (and their grammatical patterns) likely to follow the ongoing translation of a source text, and adapts to the user’s choices. Example responses of TransAhead to the source sentence “我們在完成這筆 交易上扮演重要角色” and two partial translations are shown in Figure 1. The responses include text and grammatical patterns (in all-cap labels representing parts-of-speech). TransAhead determines and displays the probable subsequent grammatical constructions and partial translations in the form of parts-of-speech and words (e.g., “IN[in] VBG[close,…]” for keywords “play role” where lexical items in square brackets are lemmas of potential translations) in a pop-up. TransAhead learns these constructs and translations during training. At run-time, TransAhead starts with a source sentence, and iterates with the user, making predictions on the grammar patterns and lexical translations, while adapting to the user’s translation choices to resolve ambiguities in the source sentence related to word segmentation and word sense. In our prototype, TransAhead mediates between users and suggestion modules to translation quality and productivity. 2 Related Work Computer Assisted Translation (CAT) has been an area of active research. We focus on offering suggestions during the translation …

Content truncated to 500 words