Meet Chao-Hong Liu, Iconic’s new MT Scientist

Chao-Hong Liu

Meet Chao-Hong Liu, Iconic’s new MT Scientist

Developments in Neural MT continue to play a crucial role in the advancement of machine translation applications. Iconic’s team of expert scientists, who develop our best-in-class Neural Machine Translation solutions, is expanding as part of this growth opportunity in language technology. Our new MT scientist, Chao-Hong Liu, is Taiwanese, and will join our Dublin office. 

We met up with Chao-Hong to get to know him better and learn about his new role at Iconic.

Hi Chao-Hong, welcome to Iconic! Could you tell us a bit about yourself?

Hello, my journey to language technology started from volunteering as a translator, when I founded a Chinese translators’ team for the GNU project (a collaborative initiative for the development of free software). My PhD research focuses on error correction of non-native sentences and speech recognition outputs, building systems that could detect erroneous segments in the sentences. I worked in both industry and academia for research and development in the areas of natural language processing (NLP) and machine translation (MT). In recent years, I have lead-organized several workshops and shared tasks on the topics of MT for low-resource languages, word and morpheme segmentation (NLP), and customer feedback analysis (language understanding). I also serve as guest editor for an MT Journal special issue on low-resource languages. 

Could you tell us about your new role at Iconic?

My role is to research and build MT systems with specific translation requirements from diverse business backgrounds, as an MT scientist. I’m also very happy to get involved in the development of systems not only for building MT systems for some language pairs and domains, but also for integrating separate systems for production purposes.

What is your favourite aspect of being a Machine Translation Scientist?

My favourite thing about being an MT scientist is that I’m working on one of the most difficult problems in both science and engineering. The problem is so difficult that many of the solutions proposed for it became very useful tools for other NLP tasks; It is the frontier of AI research. Furthermore, the problem itself is dynamic, because human languages are evolving every single day. It’s intriguing to me, to realise that the work will eventually help many people in the world.

What attracted you to Iconic?

It’s fair to say that MT architectures are mostly developed in academic environments. However, these are general MT tools and there is a huge gap between using these architectures to train MT models, and building MT systems that meet language translation needs where human translation is either limited or very expensive. It’s fascinating to have a close look at what the gap is, in terms of translation quality for businesses, and to build systems to fill the gap.

What is it like, culturally, to live in Ireland as a Taiwanese?

There are many similarities between the two nations. It is interesting to see it in terms of languages. In Taiwan, most people could speak Hoklo (or Taiwanese as we call it), but officially Mandarin Chinese is taught in the education system. Therefore, although Taiwanese is widely spoken, most people couldn’t read or write it. The Mandarin Chinese taught at school is the traditional one, which means that we have the most complex writing system in the world. There are more than five thousand ‘commonly’ used characters, compared to 26 in English. In Ireland, on the contrary, most people were educated to speak, read and write in Irish at school, but it seems that not many people actually use the language in their daily lives.

And finally, what are your hobbies, interests and passions?

I enjoy listening to music, especially music from the 1980’s. I enjoy spending time with my young son – searching for toys for him, taking him out for walks, trying to understand what he is thinking, and observing how he acquires language skills. Also, I really enjoy coding. I especially like when I’ve coded to fix a complicated problem and it then works as expected!

Welcome to the team Chao-Hong!