* Title of the invited talk

Translation between Chinese and Over 300 Other Languages: Methods and Systems

* Abstract of the invited talk

Large-scale, multilingual machine translation has long been challenging in natural language processing. In this talk I will present our efforts in developing NiuTrans - a machine translation system supporting translation between Chinese and over 300 other languages. I will discuss some key techniques used in doing so, including efficient training and inference for neural machine translation, universal multilingual modeling, Chinese-oriented improvements to the models. Moreover, I will discuss methods which make use of large language models to give a further boost to the performance.

* In-person or virtual talk

In-person

* Biography

Tong Xiao received his Ph.D. from Northeastern University (NEU), China, and joined NEU after graduation. He is now a professor of computer science at NEU, and the CEO of NiuTrans. Tong is the leader of the Natural Language Processing Lab at NEU, and the PI of several research projects on machine translation and language modeling. He is the co-leader of the NiuTrans project which aims at developing cutting-edge techniques and systems for large-scale machine translation. The NiuTrans system has been one of the most widely-used machine translation systems in China, covering translation over 300 languages. Tong is the co-author of tens of papers in top-tier conferences and journals, and the co-author with Jingbo Zhu of the widely-used textbook in China “Machine Translation: Foundations and Models”. Tong is the recipient of several rewards (e.g., the Young Outstanding Scientist Award of CCF-NLPCC).