本系消息

漢英機器翻譯系統研究與開發漫談

北京大學中文系
詹衛東

從二十世紀四十年代電子計算機問世開始，利用記算機來完成翻譯活動，就一是人們力圖實現的一個夢想。儘管潮起潮落，這個夢想的實現時而被一些人吹噓為指日可待，時而又被一些人貼上「絕無可能」的封條打入冷宮。但歷史發展到今天，人們已經能夠比較客觀公正地看待機器翻譯的研究，及其存在與發展的價值。而且，隨著全球信息化步伐的加快，從專業人士到普通用戶，都對機器翻譯投以越來越多的關注，也有越來越多的研究人員涉足這一領域，翻譯的對象也拓展到越來越多的語言對。這些，都應該算是機器翻譯開始走向成熟的最好明証吧。

作為全球最早開始機器翻譯研究的國家之一，中國跟世界其他國家的機器翻譯研究的一般發展歷程相似，也是從由外國語到本國語的機譯系統開始，繼而再發展出由本國語到外國語的機譯系統。二十世紀八十年代末、九十年代初以來，隨著一批外漢機譯系統從實驗室走向市場，漢外機器翻譯，尤其是漢英機器翻譯系統的研究工作，也在中國科學院、北京大學、哈爾濱工業大學等高校科研機構，以及中軟、桑夏等軟件公司逐漸開展起來。九十年代末，有的漢英機譯系統也開始有初期版本推向市場 ( 比如中軟公司的譯星 TranStar'99 套裝機器翻譯軟件，就包含了一個漢英翻譯模塊 ) 。

應該說，跟早期的外漢機器翻譯系統的開發相比，漢英機器翻譯系統的研發是有不少有利條件的。首先，機器翻譯的理論和技術都過了近二三十年的發展進步，人們對機譯的認識已相對清晰，而且開發機器翻譯系統的實際經驗也有了很多積累；其次，計算機軟硬件技術一日千里的發展速度，使得機譯的開發平台也非昔日可比，研發人員有更多的選擇和開發空間可供自由馳騁。再次，語言學研究和形式語法理論在二十世紀八十年代以來的發展，為自然語言的計算機處理提供了更多的更多的語言學支撐。在這樣的大背景下，現有的漢英機譯系統一般都採用相對比較成熟的機譯技術，即基於語言學規則的轉換式技術來實現自動翻譯過程。儘管不同的漢英機譯系統在具體實現時，在算法、程序以及語言知識表示的細節方面會有差別，但自動翻譯系統的基本流程大同小異。比如，一個典型的漢英機器翻譯系統，通常要經過這樣一系列步驟，首先要經過漢語的自動分詞 (segmentation) 與詞性識別 (part-of-speech tagging) ，然後對以詞串形式排列的句子進行自動句法分析 (parsing) ，得到漢語句子的結構樹 (structure tree) ，再將漢語的句法樹轉換成相應的英語句法樹，隨後根據已有的時態、性、數等句法信息對英語譯詞進行詞形變化，最終產生英語譯文句子結果。值得指出的是，上述過程中所需要的漢語分詞和詞性識別技術、漢語句法分析技術、英語生成技術等，中國學者自二十世紀八十年代中期以來都已做了不少規模較大的深度研究，特別是在一些基礎性的工作上進行了不懈努力 ( 比如漢語語法信息詞典的建設，漢語短語功能分類的研究等等 ) 。毫無疑問，這些工作都為漢英機器翻譯系統的研發打下了一個比較堅固的基礎。

不過我們也不能不正視這樣一個現實，即與英漢機器翻譯相比，漢英機器翻譯系統的質量仍略遜一籌，離真正進入實用還有一定距離。實際上這背後的原因也很自然平常。從上述我們對漢英機譯過程的諸多環節的簡要描述就不難了解，整個機譯過程中間涉及到的每一個環節得以順利通過的保証，是機譯系統要有足夠的、準確的語言知識的支持。這一點對英漢機器翻譯也好，對漢英機器翻譯也好，都是一樣的。而特別針對漢英機器翻譯系統來講，英語譯文生成質量的好壞，很大程度上取決於漢語原文分析的深度如何。要想最終產生的英語譯文句子結果，能夠準確傳達漢語原句的意思並且可讀性好，首先就要求對漢語句子的分析達到一個較高的水平 ( 比如，除了漢語句子的句法結構分析正確外，還要分析出句中謂語動詞跟前後名詞之間的語義關係，以及漢語句子所負載的時態信息等 ) ，這樣才能提供英語譯文句子生成時所必需的各種信息。但眾所周知，相比英語等印歐語法研究而言，漢語的語法研究還有很大的改進提高的餘地。在中國幾代學者的努力下，目前漢語詞法分析、短語結構的研究可以說已經有了相當的積累，達到了一個較高的水平，但漢語句子的分析，包括句子結構模式，特別是長句的結構模式，以及句型與其表意模式之間的關聯方面的研究，都還處在起步階段。這種情況，使目前的漢英機器翻譯系統，在處理漢語簡單句的英譯時尚顯應付有度，但在碰到真實語料中大量出現的長流水句 ( 結構鬆散，意合連結而成的句子 ) ，以及帶有濃厚色彩的句群的英譯問題時，不可避免地落入了左支右絀的尷尬窘境。

On Translating The Story of the Stone: A Historical Approach

John Minford

Summary

The history of translation is a much neglected field of study. Quite apart from being a fascinating study in itself, it also provides useful lessons for the aspiring tranlator. Often more can be learned from the experiences of practicing translators than from theoretical observations.

It is an opportune moment to look at the history of translating The Story of the Stone, for two reasons. The first is that David Hawkes' personal notebooks, compiled while he was translating the first eight chapters of the novel in the 1970s and 1980s, are about to published by the Translation Centre of Lingnan University. The second is the completion of an M. Phil. Dissertation at Hong Kong Polytechnic University by Chan Oisum, on this very subject. Entitled “The Story of the Stone's Journey to the West”, it chronicles the passage of this masterpiece of Chinese literature into the English language, beginning with the earliest attempt (a manuscript, hitherto unknown draft by the pioneer missionary Robert Morrison), and proceeding through the 19 th and 20 th centuries to the two complete versions by Hawkes/Minford and the Yangs.

One of the most interesting items in Ms Chan's dissertation is an appendix containing a full-length interview with Professor Hawkes, conducted in Oxford in the winter of 1998. From this interview we are able to piece together the story of how David Hawkes first encountered the novel, as a young student at Peking University in 1947; how he began to translate fragments, some of which were incorporated into his Oxford Inaugural Lecture of 1958; and how the ambitious project to translate it in its entirety for Penguin Classics was eventually conceived.

This is an instructive story for the student of the art of translation. In the interview, Hawkes specifically disowns any broad principles or general theory, emphasizing instead that he was searching for solutions to a particular novel and the problems it presented for the translator. There are many fascinating aspects of his approach to the translation, but two stand out. First, it is clear that he was from the outset in love with this book. It took hold of him right from the start. Indeed, he first read it in quite extraordinary circumstances, as a way to learn modern spoken Chinese. He sat down in Peking with an elderly retired gentleman from Hebei , and read it through. At first he understood virtually nothing, either of the text itself, or of the gentleman's explanations. It was, as himself says, a sort of direct method gone mad. The second thing that stands out, as he himself makes quite clear, is that he was seeking to create a version not for scholars but for readers of literature. This latter set the agenda; the first fuelled, motivated the invention with which he discovered solutions.

Apart from the story of Hawkes' Stone, there are other poignant and instructive episodes in this history. One of these is the fine translation begun by Maurice Bowras' great-grandfather, E. C. Bowra, who worked for the Customs Service before his premature death. The eight chapters he left augured well. Indeed, in many respects his approach reveals the same literary instincts as the later approach of Hawkes.

Herbert Giles' 25-page synopsis from the late 19 th century is an equally interesting milestone in this journey. Giles was one of the first scholars of Chinese to bring to his work a genuinely literary appreciation of what he was reading. His condensation of the story is still today a useful brief to put in the hands of students.

The later condensation of C. C. Wang, while a brave venture, suffered from its initial conceptualization. By reducing The Dream (as it was then universally known) to nothing more than a triangular love story, Wang was certainly making the book manageable. But at the same time he was robbing it of its stature. Readers (coming to this translation with their heads full of the idea that they were about to read one of the great masterpieces of world literature) were quite understandably unable to see what all the fuss was about.

Music from Two Rooms: The Lineage of Translating Chinese Literature

John Minford

Summary

The impending publication of the first volume of the large-scale anthology of English translations from Classical Chinese Literature, edited by Joseph Lau and myself for Chinese University and Columbia University Presses, is the occasion for this brief retrospect. During the 16 years of reading and research that has gone into this first volume, conducted in libraries in Europe , Asia , the United States and Australia , one of the things that we have discovered is that many of the earlier translations remain among the most civilized. The newest by no means turn out to be the best. The sober lesson to be drawn from this is that the increasing specialisation that we find in modern Chinese studies has not necessarily created good translations of literature. On the contrary, it was the broad cultural education of all men and women of letters in the 17 th , 18 th , 19 th and early 20 th centuries, that enabled these earlier and less specialized translators to come to the task better equipped. So much for the myth of professionalism.

The Point of Theory

Theo Hermans

The questions were bound to arise, and they did, more than once. When I received the tremendously flattering invitation to spend five weeks as a “Distinguished Humanities Professor” in the Department of Translation at The Chinese University of Hong Kong , I proposed a number of lectures and seminars. They were duly accepted, and I duly delivered them, in March and April of this year. They dealt with such engrossing topics as norm theory, system theory, gender and postcolonial issues, the hermeneutic and poststructuralist angle on translation, together with chapters in the history of Western thinking about translation, from the venerable patron saint of translators, the irascible Saint Jerome , to the contemporary enfant terrible, the elusive Jacques Derrida. The same questions were asked, politely but repeatedly, by students and staff, following lectures, during tutorials, over dinner: what's it all about, Professor Hermans? What's the point? What can we do with all this? How will it be of use to us, as students or teachers of translation, as prospective or practicing translators? Or, phrased more generally: why translation theory at all? Where does it stand in relation to the practice of translation?

They are natural and necessary questions to ask. They have been asked before. Other fields of study have fielded similar questions. Will linguists help us to speak or write more effectively? Does literary theory make for more successful novelists? Will the philosophy of history produce better historians?

Basically, I think there are three answers to questions about the point of translation theory. Firstly, the study of translation is a mansion with many rooms, and we should not all try to occupy the same room. Few people doubt that linguistics, and literary theory, and philosophy of history, are entitled to a place in the sun. When it comes to translation, there tends to be less tolerance. Here the assumption often is that if something cannot be directly applied to improving translation practice or teaching, it's irrelevant. I beg to differ. True, a great deal of translation theory or history is not designed for immediate practical application. It assumes instead that translation is a rich and complex phenomenon deserving sustained, serious attention in its own right. It seeks to account for what happens when translation takes place, i.e. what is involved in translating, in psychological, social, cultural and historical terms. It tries to figure out what “translation” has actually meant to different people in different places at different times. Of course, if your exclusive concern is with developing the routines and drills of translating-at-so-many-words-per-minute, you may not have time for those airy-fairy speculations. I understand that. We can still be friends, even if we don't have very much to say to each other. I will be disappointed, though, because it seems to me there is more to translation than words-per-minute or money-making and problem-solving routines. Translation is a lively, many-faceted, resplendent thing in itself. Reducing it to mechanical practicalities also diminishes it. That is why I claim legitimacy for theoretical pursuits--- especially in a university context. That brings me to my second point.

Translation theory, like translation history for that matter, is interesting, fascinating even. Yes, it can be pretty abstract. Some of it is rather remote, some of it technical --- but not more so than, say, the theoretical discourses of psychology, or sociology, or anthropology. Translation theory is interesting because it attempts to go beyond the dull immediacy of routines and deadlines to ask what happens in the brain when we translate, how languages can be mapped on one another, where current ideas about translation come from, why certain cultures or groups translate more than others, or differently from others, what criteria there might be for judging translations “good” or “bad” and who owns those criteria, and so on. There are no simple answers to these questions, but they are important, and real. They help us to understand something of the world we live in. If translation matters at all, if there is an economic, social, intellectual and personal need for it, then it is worth trying to understand it and allowing yourself to be intrigued by exactly what this thing called “translation” is, how it works, and how it can be investigated. Before you know it, you're into problems of definition and methodology, and the realization that there's more to translation than meets the eye. Translation theory is interesting because it takes us into sociology, psychology, politics, questions of identity and difference, relations between ourselves and others. The insights gained from this kind of pursuit are valuable in their own right, as part of the human endeavour to make sense of the world. The pursuit is useful for that reason. It may also serve another, more practical purpose, which leads to my third point.

Translation theory, like translation history, has definite benefits for the practitioner and the learner. These benefits are indirect, for the reason mentioned above: the theory of translation is not meant to facilitate the practice of translation, it has its own goal, which is to gain insight and understanding. It is knowledge about translation. Indirectly, however, this knowledge can benefit the practitioner, and especially the learner. I think it can be of benefit in two ways, broadly speaking.

For the first of these, foreign-language learning provides an instructive parallel. Those involved in foreign-language learning have come to realize that the so-called communicative approach, which was in vogue some years ago and focused exclusively on practical language use, can be profitably supplemented with a structural approach which tells learners about morphemes, grammar and social conventions. Methods which combine the two forms of knowledge, knowledge how-to and knowledge about, produce better results than either method applied by itself. In the same way, students learning to translate can benefit from reflecting about the choices they make while translating, the value judgments they – and their teachers – apply to translations, the range of possible solutions available in principle and in practice in any given situation, the sort of solutions that have been applied in the past with more or less success, the ways in which translators have justified particular styles of translating, etc. This self-reflexive mode of learning is more effective than learning by rote. It amounts to “deep” learning, learning that sticks because the learner has thought the matter though. It also produces more self-conscious, self-assured translators. In a word, it is not only more efficient as a method, it also empowers translators because they know why they do what they do, and can talk about it.