Will machine translation become the next wave of industry

The author of this article: Dr. Zhang Chunliang, marketing Director of Hangzhou ya Tuo Network Technology Co., Ltd., engaged in machine translation research and machine translation product development for 10 years.

Do you remember the sci-fi movie Star Trek: In Enterprise, there is a scenario in which the Starship Communications commander, Sato, invented the Linguacode matrix using the translator, and by integrating the translator into the communication pins worn by Starfleet personnel or implanted into the ear, people who speak different languages can communicate freely. In fact, the core technology of such equipment display is machine translation (Machine translation) technology. A person who used Google translation to travel around the world, using machine translation tools to communicate with locals, has also deliberately mentioned how he used the translation tool to communicate with doctors when he was hospitalized.

According to statistics, there are about 6,000 languages in the world, the story of the Tower of Babel in the Bible provides us with an explanation: God does not want to let mankind succeed in building the Tower of Babel, so spread people around the world, let people speak different languages, causing everyone to be unable to communicate freely. Machine translation technology will be the key to solving the problem if you want to rebuild the Tower of Babel and allow people with so many languages to cross the language barrier.

A mention of machine translation, we are more familiar with the Google translation and Baidu translation based on the Internet public cloud model to provide online translator services. In 2013, Google's I/O conference, Google translation team in charge of front-end products and mobile technology, Josh Estelle has mentioned that Google translation daily for more than 200 million users around the world to provide services, daily translation of up to 1 billion times a day to deal with the number of words, equivalent to 1 million books, More than the world's professional translators can translate in a year, 92% of the translation needs are from outside the United States. In addition, according to Baidu Internal statistics, the current online translation contributed more than 5% of the user traffic.

Some people call machine translation The Crown jewel in the field of natural language processing. In fact, the history of machine translation is almost as long as the development of computers. In 1949, Warren Weaver, head of the Natural science Department of the US Rockwell Foundation, published a memo on translation, marking the beginning of a historic page in machine translation research. In particular, in the last more than 10 years, the rapid rise of statistical machine translation technology, through a large number of bilingual data to automatically learn translation knowledge, a translation system can be completed in 35 days or even a day. Therefore, the translation products with the core of statistical machine translation technology are more and more favored by industry. From the perspective of translation performance, the performance of statistical machine translation system is better than that of other technologies.

With the rapid development of science and technology and social economy, the interconnection of the world has become an irresistible trend of development, the first is the deepening of the cross-language information connectivity, we are at an unprecedented speed from the single Language World (domestic) to the multilingual World (internationalization). Small to travel abroad translation or technical data translation, large to internet-based online electric business, tourism, foreign trade activities, O2O, car rental, financial services such as international marketing, an unavoidable pain point lies in the face of multilingual language translation problems. How can the massive and dynamic multilingual information on the network communicate effectively through Low-cost and fast translation? It is inconceivable to solve this problem based on human translation, the only solution is to make full use of machine translation technology to provide intelligent automatic translation services.

In the past more than 10 years, with the strong support of governments and industry, machine translation research and industrialization have made rapid progress. The Government of the United States, as early as 2009, has made automatic translation one of the ten most influential technologies in the future, and the Chinese government has included natural language understanding research including machine translation technology in the national medium and long term science and technology development Planning program. It is also to see the great industrial value of machine translation, major Internet companies are actively engaged in machine translation technology and product development.

I have with a number of professional translators to the application of machine translation for the topic of communication, five years ago, the feedback was basically "sing decline" machine translation negative view, that machine translation can not be compared with human translation. However, in the past two years, the translation industry, especially the translation company, has a great positive change in the evaluation of machine translation, gradually began to accept the availability of machine translation conclusions, and even a large number of people are afraid of machine translation, fear of losing their jobs. Judging from the maturity of the current machine translation technology, it is not possible to replace the manual translation in the foreseeable future, but it is possible to replace some low-level translators. In fact, if you build a professional machine translation system for a specific area, it will be very usable, providing very good assistance for translation services and 4 million of professional translators in our country.

In real-time speech translation, Google, Baidu and other internet giants using speech recognition/synthesis and machine translation technology developed a mobile phone version of the Voice translation app. As long as you carry the mobile phone with the app, you can communicate freely in different languages, which is especially suitable for the general public to travel abroad, truly "one machine in hand, all over the world." In a demonstration of Skype's real-time translation technology in the May 2014 Code Conference, Microsoft vice President Pal made a conversation via Skype using real-time translation technology with another Microsoft employee, the former in English, the latter using German, and the demo content showing Skype successfully completes the real-time translation of both dialogues.

The 2014 "Double 11" shopping carnival, on the Alibaba platform, the global total of 217 countries and regions involved in the snapping. Not only is the Ali department, to do "Cross-border electricity dealers" to "buy and sell all over the world" has become the consensus of all electricity dealers. By using machine translation technology to solve the cross language communication problems of buyers, sellers and suppliers of platform providers, ebay is at the forefront. Ebay has tried machine translation when it opened up the Russian market in 2013, according to TechCrunch, a technology blog. Russian users can also return items that match the keyword and are described in English when they enter a Russian search keyword. Alibaba in China also has its own machine translation research team, has developed a number of English and Russian, English and Portuguese between the multiple sets of machine translation system. Machine translation technology can also make a difference in other aspects of the service, such as the multi-Mandarin machine translation Cloud Platform and web-based online service of the artificial customer service system, can provide multilingual intelligent automatic translation services, to help customer service staff and foreign customers for friendly exchanges.

If you combine the translation cloud platform with the chat platform, we can achieve friendly communication based on text/voice/video from friends from different countries, which can provide language communication service for social platform, online gaming platform and so on, such as the current micro-letter is to provide Chinese-English, Hande and other 21 languages of machine translation services. Today, smart equipment industrialization is very hot, with integrated speech recognition/synthesis of wearable equipment and smart chip similar to integrated machine translation technology wearable equipment and smart chip, with the 4G communication technology mature, I believe that in the future there will be a cool space for imagination.

In addition, in the current booming online education market, including MOOC, a variety of large-scale open network curriculum platform by the tens of millions of of learners, but for a Chinese-speaking learners, how can the online education platform to listen to the insights from Professor Harvard? Through machine translation technology, we can share the world's best teaching resources to hundreds of millions of learners around the world, so that everyone transcends the language barrier, truly have equal, the best education opportunities.

In addition, we should point out that China is a unified multi-ethnic country with many languages and many kinds of texts. The work of the national language is the important content of consolidating the socialist national relations and carrying out the national work, and developing the multilingual intelligent Translation system for minority languages can further promote the application of the minority languages in the fields of bilingual teaching, adult literacy, press and publication, Internet and so on, to stabilize the nation and promote the economic Political construction, cultural construction and social construction have far-reaching significance, so the country is giving a lot of human and financial support.

Among the various machine translation applications mentioned above, the core part is the construction of multi-Mandarin machine translation Cloud platform. The simplest solution is to provide online translation services based on the Internet public cloud model, such as Google Translations, Bing translations and Baidu translations. By using the online translation API for two development, you can basically meet the popular use of machine translation and lower requirements of enterprise applications, but only if the amount of translations can not be too large (many API translation requests are limited), the results of translation requirements can not be too high (the above system for the general domain translation, Personalized custom service is not supported. One of the biggest practical problems is the security of data translation, which is a matter of great concern to many international enterprises, government departments and national Security Intelligence. If translation is in great demand and the security level required for data translation is high, the better solution is to help users break through the local multilingual machine translation private cloud platform. To this end, many of the companies that need to build their own translation teams to use open source systems such as Moses and Niutrans to build machine translation private cloud platform; For other enterprise users, SDL's enterprise-class translation servers and Niutrans Server can provide specialized multilingual machine translation private cloud services.

To sum up, we have good reason to believe that machine translation technology is changing people's work and life style in the development trend of the single language world and the multilingual world. Perhaps 35 years later, now in many people's eyes still belong to "tall" machine translation will go down the altar, and constantly into everyone's daily life, the benefit of society, machine translation industry will bring a new wave of industry.

(Responsible editor: Mengyishan)