Born – Colombo, Sri Lanka
PhD – University of Tokyo, Japan
Joined University of Liverpool – 2013
Position – Professor in Natural Language Processing, Department of Computer Science
What is your research about?
I am a researcher in Natural Language Processing (NLP), which is a branch of Artificial Intelligence where the goal is to make computers understand and process languages spoken by humans. NLP has many applications such as information retrieval such as search engines, dialogue systems such as Alexa and Siri, machine translation, text summarisation, and question answering, amongst others.
What or who first inspired you to be interested in your research subject?
As an undergraduate student, I read a paper by Professor Mirella Lapata on how to order sentences in a summary to make it more coherent. I always loved maths and was impressed by how probabilities could be used to model human languages as done in that paper. I also loved learning languages and was studying Japanese at that time in University of Tokyo. Therefore, I thought NLP was the right field of research for me.
What are you most proud of achieving during your research career so far?
There are many, so it is difficult to pick one. I invented a method to accurately measure the similarity between two words using web search engines. It can handle words that are not in human-compiled dictionaries and can be used for any language. This idea has sparked lots of follow-up work and I received the IEEE Young Author Award for it.
What techniques and equipment do you use to conduct your research?
NLP algorithms are implemented using programming languages such as Python and Java. They typically do require large amounts of textual data and processing power. Therefore, I use cloud-based GPU-servers in my research.
Which other subjects are important for your research?
Mathematics is very important in my research, especially linear algebra, calculus and probability theory. Machine learning and some linguistics preliminaries are also important too.
What is the key to running a successful research group?
When a research group becomes large it is difficult to supervise each student on a 1:1 basis. It introduces an unnecessary bottleneck and slows down the progress of research. Therefore, I try to organise students into teams that share common research themes and interests. That way they can help each other, learn from each other, and conduct research more efficiently. I try to connect senior PhD students and Research Associates with junior students.
What impact is your research having outside of academia?
NLP is a highly applied field with lots of interest from industry. I have always maintained a strong connection to the industrial applications of NLP. Currently, I work for Amazon as an Amazon Scholar and contribute to information retrieval and extraction technologies used in Amazon product search.
How do you plan to develop your research in the future?
In my dual role as a university-academic and an industrial-researcher, I am interested in identifying and modelling real-world problems in industry that can be formalised into long-term research problems that could be of interest also in an academic setting. I would like to develop my research by solving such problems.
What problem would you like to solve in the next 10 years through your research?
Given the wide applications of NLP systems in the real world by billions of users, I consider problems related to social biases of NLP systems to be an important problem that must be solved. Moreover, I am interested in unsupervised methods to train NLP systems such that they can provide explanations to their actions. I have already started working on some of these problems but a lot more need to be done.
What advice would you give to someone considering a career in research?
To come up with good ideas, you must not fear trying out many ideas. It is important to read a lot of related and relevant research papers to see the connections and to come up with novel ideas. Then, discuss your ideas with your colleagues, friends, collaborators, and supervisors.
It is also very important to write your ideas down, as writing makes it easier to see the flaws in your ideas. The ability to build prototypes quickly and conduct experiments rapidly will help you to progress your research. In computer science this is relatively easy and inexpensive because if you are a good programmer then you can quickly implement your ideas and test them. If you need to modify your idea, all you need is to modify your code and re-run the experiments. You can also test many ideas in parallel in computer science, which is remarkable if you think about the process of experimentation in other fields.
Back to: Faculty of Science and Engineering