previous arrowprevious arrow
next arrownext arrow
Shadow
Slider

Introduction to Professor Lun-Wei Ku and Her Lab at Institute of Information Science, Academia Sinica

  1. Could you briefly introduce yourself (and your University/Lab)?

I am now an associate research fellow in Institute of Information Science, Academia Sinica, and an adjunct associate professor of the National Chiao Tung University (NCTU). I received my M.S. and Ph.D. degrees from Department of Computer Science and Information Engineering, National Taiwan University. Before my academic career, I have been working in industry at Acer and IntoVoice for speech recognition. My research interests span over natural language processing, information retrieval, and computational linguistics, especially on subjective information analysis such as sentiment analysis and opinion mining, emotion detection and stance classification, and NLP applications. I have been working on Chinese sentiment analysis since year 2005, and previously was the co-organizer of NTCIR MOAT Task (Multilingual Opinion Analysis Task, traditional Chinese side) from year 2006 to 2010. Now I organize the SocialNLP workshop collocated with top conferences every year, where this year is its 10th anniversary. I participate actively in the research community, too. For example, I was the secretary-general of Association for Computational Linguistics and Chinese Language Processing for 6 years. Recent professional international activities I involved include serving as general chair, program chair, best paper committee member, area co-chair in top conferences, member-at-large in Asia Federation of Natural Language Processing, and information officer (Taiwan Area) in ACM SIGHAN. With all the research achievements, I was recognized by CyberLink Technical Elite Fellowship, IBM Ph.D. Fellowship, ROCLING Doctorial Dissertation Distinction Award (2009), and Good Design Award Selected. In the meantime, our lab publishes papers in top conferences such as ACL, NAACL, EMNLP, WWW, SIGIR every year. I am also experienced in industrial collaborations. Selected partners include HTC, banks, web data, and medical information companies. These collaborations create the deploying opportunities of the technology developed from my lab, as well as job opportunities for my students. More details can be found on my professional website

My research lab, the NLPSA Lab (Natural Language Processing and Sentiment Analysis Lab) is dedicated to the research on natural language processing, especially on subjective information processing. The lab members are mostly students in collaborative or international programs from top Taiwan universities plus students who plan to pursue their M.S. or Ph.D. degree abroad. In addition, we collaborate with other universities worldwide. In sum, NLPSA is a comfortable, friendly and international working environment equipped with high-performance hardware and GPUs. More details can be found on my lab website.

My research institution, Academia Sinica, is the most preeminent academic institution in Taiwan. There are a total of 32 institutes and centers in the mathematics and physical sciences, life sciences and humanities and social sciences divisions. You can visit here to know more.

  1. What have been your most significant research contributions up to now?

My research interests lie in subjective information processing, specifically sentiment analysis and opinion mining, which is a subarea of natural language processing (NLP) and its interplay with knowledge inference (AI), computer human interaction (CHI), recommendation systems (IR), and computer assisted language learning (CALL). Sentiment analysis and opinion mining is a science about how to understand the sentiment information expressed by people or aroused by content, especially through text media. Although sentiment analysis and opinion mining is a comparably young research subarea of NLP, related technologies have advanced quickly, making it one of the most competitive subareas in NLP thanks to the dedication of a vast number of researchers. Moreover, due to its cross-discipline nature and its real-world applications, new and exciting directions continue to emerge to address the general goal of understanding. I have great interest in proposing and developing such exciting research directions. For example, text mining and recommendation on social media, which relies heavily on understanding the sentiment of users and their posts, requires effective features that describe users and posts for better profiling. I am interested in generating article and user representations from the aspect of sentiment context and its engagement to users. In a different direction, I have recently worked on how to utilize subjective information analysis techniques to improve the quality of life by detecting people’s subjective information (e.g., stances, emotions, deceptions) as well as by helping people sense such information from other individuals in the same environment; in contrast, much other research ceases after relevant techniques have improved to a certain level. I believe this direction will naturally lead to the most appropriate design for dialogue systems and text generation systems for various genre, which is currently in strong demand by industry. I have also worked on foundational topics in natural language understanding such as lexical and sentence inference and knowledge graph utilization in hopes of extending the advances of sentiment analysis and opinion mining by better understanding languages, and hence drawing nearer to a solution of the challenging inference problem. The well-known research from my lab includes NTUSD (ANTUSD), CSentiPackage, EmotionLines, EmotionGIF, and Reactive Supervison (SPIRS).

  1. What problems in your research field deserve more attention (or what problems will you like to solve) in the next few years, and why?

Similar to many NLP labs, most of our current and near-future research projects focus on using deep-learning approaches in natural language processing. However, we are mostly interested in how to bring human knowledge and learning ability into the current deep learning paradigm so as to enable the ability of the model to work under the few-shot or the zero-shot condition. In addition, we are passionate about encouraging people understand and accept information or decision provided by AI models. The key problem we try to solve is to create a smooth cooperative working loop between models and human so that the AI technology can really contribute to our life.

  1. What advice would you like to give to the young generation of researchers/engineers?

Patience is the best friend of a successful research. Keep trying and take comments as well as suggestions. It is always good to learn from other disciplinaries and consider the relation between technology and humanity. At the end, let me share words I always use to encourage myself: “Have the courage to be different! Be yourself!”