双语学通 | 著名计算语言学家冯志伟教授来外国语言文化学院谈“机器翻译与人工智能”

发布单位:台网学术编辑:张艺萌 记者:郭天也发布日期:2018/12/29浏览量:62

1228日下午,我国著名计算语言学家冯志伟教授到访外国语言文化学院,在33号楼211室做题为机器翻译与人工智能的讲座。外国语言文化学院院长李佐文、副院长刘颖出席讲座,我院教师与研究生共40余名参加了此次讲座,讲座由外国语言文化学院严玲老师主持。


冯志伟教授是国内唯一一位获得自然语言处理杰出贡献奖的学者,很早就开始研究数学与语言的关系,是计算语言领域的开拓者。冯教授曾使用汉语、德语、英语、法语等语言发表专著,硕果累累。

通过目前学界关于计算机是否会淘汰人工翻译的讨论,冯志伟教授指出了国内翻译专业面临的困境。但冯教授同时表示,目前来看,外语教学仍然是不可替代的。他指出:我们外语人应该适应社会的需要,同时掌握语言和技术。



西方文化中有关于巴别塔的传说。而冯志伟教授指出,中国在周朝时期也出现了承担翻译责任的象胥。如今,在信息时代,人类为了克服由语言差异带来的交流障碍,曾有人提出使用人类通用语言(lingua franca来代替不同语言的想法。而冯志伟教授指出,这种想法是难以实现的:不同语言代表不同的文化,如果全球使用同一种语言,那人类丰富多彩的民族文化也就黯然失色了。因此,冯教授提出,既然要使用多种语言进行沟通,对翻译服务的需求也就自然变得越来越迫切和尖锐。

接下来,冯志伟教授提出了机器翻译与人工智能同甘苦共命运的观点。冯教授指出,机器翻译可以帮助人们及时地获取信息,克服语言的障碍,重建人间的巴别塔。1954年,美国乔治敦大学在国际商用机器公司(IBM)的协同下,进行了世界上第一次机器翻译试验,这次试验的成功催生了1956年的达特茅斯会议。正是在这次会议上,麦肯锡首次提出了人工智能(Artificial Intelligence)的术语。


冯志伟教授随后对机器翻译与人工智能做了详细阐释。首先,他指出机器翻译包括三种基本类型基于短语的翻译、基于统计的翻译以及基于神经网络的翻译。随后,冯教授详细介绍了FAJRA系统的工作原理,该系统是世界上第一个汉语到多种外语的机器翻译系统。接着,冯教授展示了该系统将一段中文文本翻译为英语,法语,日语,德语的具体实例。此后,他还介绍了信息时代基于大数据分析应用的各种在线翻译服务和口语翻译器,肯定了当前基于人工智能和大数据支持的翻译成绩。但冯志伟教授同时补充道,在文学翻译方面,机器翻译的水平还有待提高。

最后,冯志伟教授同在场师生进行了讨论,就机器翻译与语言学的结合、该领域的具体发展以及翻译专业的学生如何更好地适应社会需要等问题进行了解答。


讲座后,院长会见冯教授,并就翻译专业学科建设进行会谈。


冯志伟,北京大学和中国科学技术大学双硕士,先后到法国格勒诺布尔理科医科大学、德国夫琅禾费研究院(FhG)新信息技术与通信系统研究所、德国特里尔大学语言文学院、德国康斯坦茨高等技术学院国际术语学和应用语言学中心(CiTaL)、桑夏自然语言处理研究院、韩国科学技术院(KAIST)电子工程与计算机科学系(EECS)学习、工作。现为国家教育部语言文字应用研究所研究员、博士生导师、学术委员会委员。著名计算语言学家,从事机器翻译,自然语言处理和计算语言学等基础理论研究和系统开发几十年,硕果累累,学术成就斐然。

Prof. Feng Zhiwei Delivers An Academic Speech on Machine Translation and Artificial Intelligence



On Dec. 28, 2018, Prof. Feng Zhiwei, a well-known computational linguist in China visited the School of International Studies, CUC, and gave a lecture on Machine Translation and Artificial Intelligence. The event was hosted by Prof. Yan Ling.

At the beginning of Prof. Feng's lecture, he pointed out the necessity of translation in the information era. First of all, since a nation’s language is indispensible from its culture, man shall never ditch the various existing languages for one “lingua franca” to simplify the cross-cultural communication. Nowadays, multilingualism has become a wide-spread phenomenon in many countries, thus translation still ranks high on the agenda of the communication of the information era. Futhermore, accompanied by the information explosion and the arrival of the Age of Big Data, multilingualism is also sprawling across the Internet, which makes the needs for translation grow rapidly.



Prof. Feng said that our time still needs good translation services, despite the traditional way of it not meeting the needs of today’s market. The machine translation is exactly what our time calls for now.

Introducing the notion of machine translation, he deemed the machine translation as the simulation of artificial translation. Prof. Feng also talked about the history and development of Artificial Intelligence, and the connection between machine translation and Artificial Translation.



Prof. Feng gave a detailed introduction to the three approaches to machine translation, namely Phrase-Based Machine Translation (PBMT), Statistical Machine Translation and Neural Machine Translation. According to Prof. Feng, even though the PBMT made a great breakthrough in the machine translation, it still has conspicuous flaws. Statistical Machine Translation and Neural Machine Translation, on the other hand, are both based on the big data, which deepens our insight into translation technology.

Today, most widely-used machine translation mechanics are based on pure data-driven Artificial Intelligence. Having made great contribution to the quest for a more reliable, more efficient translation service, however, they are still not free of flaws. The inspiring lecture of today not only reminded us of our hitherto achievements, but also pointed out that there is still a long way ahead to develop truly optimized machine translation applications.

The Dean of the School of International Studies, Li Zuowen talked with Prof. Feng about the disciplinary construction of translation after the speech.