Many countries in the world have been engaged in this study for a long time. In fact, since the birth of the electronic computer in the 1940s, it began to explore the computer to language translation. The study of my country's machine translation can be traced back to the mid-1950s. This year is the 40th anniversary of the successful demonstration of the first machine translation test in my country. At the test of 40 years ago, although the scale is not large, it is already the world level at the time. The country in which the trial in the world is active in the world is really flexible. Half century, the world's scope and the mechanical translation study of my country have passed a paragraph road, and there have been about 10 years of stagnation or silence after the mid-1960s, which is not the same. It is mainly the impact of an unmanned Alpac report published by the United States in 1966, has stopped support for the translation research of machine translation. The Alpac report is mainly to: After investigation, the machine translation speed is slow and accurate, which is much higher than the labor translation cost. It has not been expected to develop a practical machine translation system in the near future or foreseeable future. Although this report was later received by many serious criticisms, it was considered to have a serious prejudice, but it also had a lot of damage to machine translation research. Until the mid-1970s machine translation began to recover and be prosperous in the world. Japan is a country that is very unaffected by the world's stagnation. Almost all large computer companies in the early 1980s conducted research and development of machine translation systems, such as Fujitsu, Hitachi, Japan Electric, Toshiba, Sharp. Japan's contribution to promoting machine translation research is recognized by the world. Under its advocacy, in 1987, the first machine translation summit (MT Summit) was held in Japan, and was decided to take turns in subaranda, Europe, and US. Soon, the Asia-Pacific Machine Translation Association, the European Machine Translation Association, the North American Machine Translation Association, and the International Machine Translation Association were also issued regularly "machine translation." In September this year, the seventh summit was held in Singapore. It is also the last summit of this century. Its theme is "the machine translation of the new century translation." my country has introduced the status quo of my country's machine translation research and development at the "Development of Motionic Translations around the World", and representatives, which were invited to participate in the theme seminars in funding investment and introduced my country's various channels on machines. The investment situation of translation research. The author was invited to have a report entitled "Past, now and the future" of English and Chinese / Chinese Machine Translation in the academic seminar before the meeting.
Today's machine translation should be no very strange words for many people. Today we can buy a variety of PC machine translation software in the software store, all languages, such as English to Chinese, Chinese to English, or Japanese to Chinese, and even English to Japanese, what else The family version, the professional version, and the various different professional dictionaries are available. It is estimated that there are more than 1,000 different machine translation software in the current market in the world, and there are nearly 20 PC machine translation software with certain sizes of my country. The price of PC machine translation software in the world is not expensive, and the price is still reduced. Now we can even enjoy the service of the translation system. So now users have large choices. Of course, a user should first understand their needs very well when choosing a machine translation system. Specifically, we have the following suggestions.
First, if your translation task is more stable or long-term, professional single, the translation result is to be published, you may be a translation company, or a professional intelligence, then you can choose to have a large scale In the field of professional dictionary, and it can be derived from a large amount, long-term translation operation (some system crash) system. At the same time, it is more ideal that you can also have a "translation memory" system that can help you handle text format issues (such as font, chart, footnotes, etc.), and you can save you the modified correct translation for later translation Reuse.
Second, if your translation task is temporary, professional is not single, the translation quality requirement does not need to reach the level of publication, then you can choose a professional dictionary with multiple fields, but it should also be a large amount, long Time translation running system. Third, if you are using the online information to be translated, you must choose the system that can run online. If your foreign language is equally, it is limited, and it is also possible to choose a system with only a large-scale dictionary but can be compiled with a point.
Today, machine translation can be said to be quite prosperous than 10 years ago. But we are willing to remind, there is a crisis behind this boom. As mentioned earlier, the trauma brought to the machine translation now seems to have been well. But in fact, its shadow always will always appear on the head of the machine translation researcher from time to time. Now, with more and more translation systems to market, the government's investors feel that if they still have to invest in the investment, it seems that there is a little name. And the merchant just wants it now to make ready-made technology packaging packaging. I often hear the bosses will ask the researchers. "How long does you expect to develop products? How is your system correct?", Probably no researcher will answer, the future "The correct rate is about a percent Fifty left and right. If you answer, then his project is still "shot". However, existing machine translation systems (not only in English-Chinese or Han English, other language pairs of foreign languages), how much is it actually actually in the face of real text? The translation quality of the machine is indeed very satisfactory. Recently, some abroad, some people who have done "MT, not the abbreviation of Machine Translation, but the abbreviation of Mad Translation (translation of madman). They are almost in battle with the machine. They advised people not to buy a machine translation system, To translate translators. Some people in China said that there is a machine translation. "Full of English can't help, full of Chinese can't understand." These are extremely extreme evaluations, but the quality of machine translation is indeed It has been a difficult problem. The famous machine translation critics said in the recent machine translation summit saying that the quality of the translation translation has not achieved substantial progress, and the problem that has not been solved 50 years ago still exists. There is also a deeper crisis, which is from the researchers themselves. They say "under the existing technical conditions, the quality of the translation" can only be like this. "It seems that they are not" existing technical conditions "creator. So, the situation that may arise will be invested in investors and decentrs in a lower level of system, but no longer At sufficient funding and technical investment. Machine translation is not yet mature in theory or technology. Now because people have a strong demand for overcoming language exchange, especially the emergence of the Internet is more prominent. The machine translation has achieved a chance to meet this demand with a lower translation, and use this opportunity to seek further development. We have a clear understanding of this reality. On marketing, it should be impossible Promote. Now in the newspaper magazine, you can often see the over exaggerated propaganda of the machine translation system. From the long way, this is the unwanted "self-smashing brand". The correct practice is to get the product there, Horses tell them what the machine can do and can not do anything, how to use it, use it, avoid it short. At the same time, debug and improve the system according to the needs of users. In other words, do more Cultivate users, cultivate systems, and cultivate the market's work. For about 20 years, the method of machine translation research is really refurbished, and it is dizzying, there is rules based, based on knowledge, based on corpus, based on statistics and textbooks, based on statistics and textbooks For example, based on dialogue, from another angle, there are direct methods, conversion methods, intermediate language, and so on. Which of them has not been able to achieve substantial breakthroughs in translation quality. How can I get it? Substantial improvement? We may wish to compare existing translations and people. Motion translation:
1. One sentence is handled, the first sentence does not know what is the content of the second sentence, the second sentence, no longer refer to the first sentence;
2. The analysis of the source language is just to solve the syndrome relationship, which is not a sense of understanding;
3. Its developers ask it almost universal. It seems that it can cope with it, from the computer to medicine, from chemical industry to the law, it seems to change a professional dictionary; 4. Its translation conversion is based on The syntact structure of the source language, the syntact structure of the source language;
5. Its translation is just a mechanical correspondence of the syntactic structure and the phrase.
Man translation:
1. Generally read through full, he will take before and after;
2. The source language is the understanding of the meaning;
3. Only professional translators, no one can be hit by universal translators in the world;
4. His translation is based on his understanding of the source language, and does not accept the syntact structure of the source language;
5. His translation is a process of re-creation.
The machine translation study is a knowledge processing problem. It involves knowledge of knowledge, language between languages, and world knowledge outside the language, including common knowledge and related fields. I think that from a practical point of view, automatic high quality machine translation should not be a target, at least a near-term goal, but from the perspective of research, high quality should be a target. Because so we can not only build a machine translation system, but also explore the mechanism of man translation. In recent years, I have emphasized that machine translations should be broken in many occasions, and there is an innovation. The machine translation research in the next century should break through the following three aspects:
First, the big language, and no longer handled one sentence in a sentence;
Second, based on understanding, no longer staying in the hierarchy of syntactic analysis;
Third, highly specialized, specialized, no longer a "all-in-way, sample loose".
Professor of several mobile translation authorities in September unanimously proposed deep basic research, making the analysis of source language based on understanding. Calling on investors to increase investment in basic research, such as large-scale knowledge dictionary. After more than ten years of hard work, the author has established a large-scale knowledge system "How to know this year, this year has been used for free use. It has caused widespread attention from scholars at home and abroad, and has started new explorations on it, as described above, as well as methods based on understanding, and new exclusion ambiguity.
With the popularity of the Internet, with the advent of the information age, the application prospects for machine translation are broad. As a window that explores its own intelligence and operational mechanism, machine translation research will be more attractive. I hope that the problem that has not been solved in 50 years will be resolved in the new century.
Copyright? 1999 - 2002 Keenage.com,
All Rights Reserved
Email: support@keenage.com
Contact number: 010-82382578-205
