UM Develops the World’s First Smart Chinese-Portuguese-English Translation Platform

Chinese Text| Kelvin U
Photo|Jack Ho with some photos provided by the interviewee

With the popularisation of increasingly sophisticated machine translation systems, language difference need not be the barrier that it used to be. According to statistics from the Globalization and Localization Association, the market size of the global language services industry reached USD 45 billion in 2018. By the year 2021, this figure is expected to rise to USD 56.2billion, with the Chinese market claiming USD 5.39 billion. Seeing the growing demand for translation services worldwide, a team of researchers from the University of Macau (UM) has developed a Chinese-Portuguese-English machine translation system to assist translators and enhance the efficiency and quality of their work.

Meeting the Needs of Translation Companies

With advancements in artificial intelligence and computer technology, machine translation has become one of the key areas of research at UM. In this respect, the Natural Language Processing &Portuguese-Chinese Machine Translation Laboratory(NLP2CT) has made tremendous progress over the past two decades, as is reflected in the various innovative translation systems it has developed, most notably UMCAT, the latest online translation system that can effectively assist professional translators in the translation, proofreading, and management of translation projects. In addition to producing smart full-text translation with a high degree of accuracy, UMCAT has various other functions, such as creating industry-specific terminology databases, enabling clear division of labour for large projects, effectively monitoring translation and review progress, and suggesting terms for reference. It is suitable forindividuals, government departments and companies that need to handle a large amount of translation projects involving Chinese, English and Portuguese. Launched in December 2018, UM-CAT now has close to 100 registered institutional users, both at home and abroad.

Wong Fai, director of the NLP2CT and associate professor in the Faculty of Science and Technology (FST), explains that UM-CAT has two features that help enhance translation efciency. First, the multiple built-in functions, such as smart translation, translation memory, glossary management, collaborative translation, and AI-based prompting, enable the system to show the translations that were produced, suggested, or adopted by different translators for the current user’s reference, in order to ensure consistency and accuracy. Second, the user interface that integrates the various steps of a translation project, from translation to proofreading to typesetting, enables flexible division of labour for different projects, facilitates day-to-day management of complex projects, and allows users to effectively monitor translation progress and view the final product, thereby enhancing team efficiency. Moreover, the NLP2CT personalises UM-CAT through UM’s Centre for Innovation and Entrepreneurship to better meet the needs of companies and provide the complete solution for translators.

Professor Wong Fai’s research team

Tailor-made for the Local Market

You may wonder why anyone would need UM-CAT when there are already free online translation platforms such as Google Translate. Prof Wong explains that while Google Translate does a good job of translating generic terms, it falls short when it comes to culture-specific terms. For instance, if you type in the Chinese name of the Ruins of St Paul’s Cathedral in Google Translate, it may not produce a satisfactory translation. Another example is ‘Fai Chi Kei’, a residential area in Macao. In Chinese, ‘Fai Chi’ means ‘chopsticks’, and ‘Kei’ means ‘basics’, so Google Translate is likely to translate this term into ‘chopsticks basics’. It is an inevitable shortcoming with online translation platforms like Google Translate because developers tend to be geographically indiscriminate in the design process. But this is precisely where UM-CAT comes in. It can translate Macao’s street names, government departments, legal terminology, and other terms commonly used in Macao, with a much higher degree of accuracy.

UM-CAT has another function that is lacking in online translation platforms like Google Translate—it enables the user to revise an unsatisfactory translation produced by the system and remembers the manually revised version for future reference. Prof Wong believes that UM-CAT has a role to play in the sizeable language service market. ‘Big companies all need translation services to varying degrees, and a system like UM-CAT not only can help improve the speed and quality of translation; it can also assist in the management of old files,’ he says.

Prof Wong Fai’s team receives the Science and Technology Progress Award at the first Macao Science and Technology Awards

Will It Ever Replace Human Translators?

Whether or not you are a professional translator, it is natural to wonder if this type of technology will ever replace human translators? Prof Wong answers with a resounding ‘No’. He explains that this misconception derives from a lack of understanding of machine translation. ‘The intended function of machine translation is to realise automation of the process,’ he says. ‘In other words, the machine suggests a translation, but it is up to the translator to decide whether to accept or reject the suggestion based on his or her expertise.’ According to Prof Wong, the translation produced by the machine is not always dependable, so a translator plays the essential role of a gatekeeper in order to prevent embarrassing mistakes. For instance, the machine is likely to translate ‘男士免 進’ into ‘men’s free entry’, which is the exact opposite of the intended meaning (‘no entry for men’). Also, as Prof Wong stresses, language is always changing, so too must machine translation systems, and for that to happen, we need humans to ‘teach’ the machine. For these reasons, it is unlikely that machine translation will completely replace human translators.

Prof Wong says, ‘Over the past five years, the size of the machine translation market has nearly doubled, but on the other hand, the size of the human translation market also maintains an annual growth rate of 10 per cent. When mainland China and Macao reach the same degree of internationalisation as Europe and the United States, the demand for translation services will likely double. With competent human translators, the quality is guaranteed, but humans take longer time and have limitations as to the number of languages they can handle. Machine translation is the exact opposite. It is faster and can handle multilingual projects, but tends to produce inconsistent results. The inexorable progress of globalisation brings both opportunities and challenges for human translation and machine translation. Only by embracing the merits of both options can we achieve win-win results.’

Professor Wong Fai

The Pioneer of Machine Translation in Macao

Developing a machine translation system is in essence a process of teaching the machine to ‘understand’ and ‘contemplate’ language and acquire cognitive intelligence much like humans, which is one of the greatest challenges in the development of artificial intelligence technology. Before Macao’s return to China, research in this area was virtually nonexistent. The gap was filled in 1999 when Prof Wong developed the world’s first Chinese-Portuguese electronic dictionary, making him a pioneer in machine translation research in Macao. Soon after the launch of the e-dictionary, more than 800 government departments and private companies became registered users. ‘Before the handover of Macao’s sovereignty to China, a large number of official documents needed to be translated, but there was no software or system to aid the translation, so the development of the first e-dictionary was driven by a practical need,’ he says. In 2007, during his visit to UM, then Portuguese President José Sócrates praised the University for its Achievements in Chinese-Portuguese translation studies, which he believed would promote the friendship and collaboration between China and Portugal. His prediction has come true. The machine translation systems developed by UM over the years have indeed greatly facilitated Sino-Portuguese communication in business and cultural affairs. Prof Wong says, ‘UM has established an interdisciplinary research platform, which is focused on the studies of natural language processing, machine translation, machine learning, and text inference, as well as the development of related systems.’

Prof Wong Fai presents the latest progress of UMCAT at a joint seminar co-organised by UM and the Institute of Automation, Chinese Academy of Sciences.

Tackling the Challenge of Voice Translation

2019 marks the 20th anniversary of Macao’s return to China. The next goal for Prof Wong and his team from the NLP2CT is to tackle the challenge of Chinese-Portuguese voice translation. ‘As a spin-off of machine translation, voice translation is considered to be the most commercially promising area in the next five years,’ he says. Currently, theNLP2CT is working on developing a smart conference translation system that can accurately identify and translate the languages used in a multilingual conference to enhance the efficiency of the conference process. The system is currently being tested and is expected to be launched for the mass market in the near future.