Chih-Hao Tsai's Research Page >


1998, A Year of No Significance:

The Meaning of Life of a Psychologist

Chih-Hao Tsai ()

University of Illinois at Urbana-Champaign

The title is actually adapted from one of my favorite books by Ray Huang: 1587, A Year of No Significance: The Ming Dynasty in Decline published in 1981 (Yale University Press). Most people think there are only some big historical events, such as those presented in history textbooks. A good historian, however, pays attention to regular, ordinary events and infers their underlying meanings: the dynamic, psychological, social, cultural, and political factors that had produced those events.

1998 is my 6th year in graduate school, 10th year in psychology, 15th year in computer programming, 26th year in school, and 29th year in the world. Nothing particular. But as my life is still unfolding and the branching factor is still large, I want to know where I am going and where I really want to go. I want to know the meaning of (my) life as a psychologist.

First and most important of all, I love psychology and identify myself as a researcher in psychology. But identity does not mean happy. Research ideas keep coming up but few of them got positive feedback. This is frustrating and if the situation persists, it could make a graduate student give up.

Secondly, I really do not feel secure enough to be a pure psychologist. Theoretical psychology will not be of too much help finding me a good job outside academia, and inside academia I do not have too many job opportunities, either. In addition, pure theoretical research can hardly be understood by people outside the field. I do not want to be alone and I do not like lonely.

I could have given up and left academia to pursuit a better life. Even being paid only $1500 a month for some routine, technical work would be better than working hard on research but receive no reinforcement. In academia, the intermediate processes are not rewarded. People only pay attention to the final output (i.e., publications). Perhaps this is not a problem for a mature researcher, but it is indeed a problem for a graduate student.

I really want to be a psychologist, and I really love psychology. I need to survive.

My rescue is programming. Programming is fun. If I bring programming and psychology together, I may be able to make them support each other. Although I have been a programmer for more than half of my life, I really did not have enough professionalism. So I took courses such as Data Structures and Software Engineering. I also took some higher level courses such as Computational Linguistics.

The purposes of becoming a professional programmer are twofold: to acquire skills to implement psychological theories (i.e., computer simulation), and to acquire skills to develop commercial software.

To be able to implement your theory is important. We focus on psychological "processes", and the term "process" was borrowed from computer science. Psychologists in general are loose in using that term. Most of the time when they say "process", what they really mean is "algorithm". Although algorithms can be more or less independent from the implementation level, to understand an algorithm well, one must implement and run it. Experimentation is the way most psychologists are using, of course. Results from experimentation provide constraints to the solution of a particular problem. But my experience was that you could never know whether the constraints are sufficient unless you implement and run it.

On the other hand, as a programmer I am old. I have to carefully choose the domain that I want to devote myself into. Being a psychologist I have the advantage of understand more about the human information processing system than other programmers do. I should make good use of this advantage by writing applications that require natural language processing capabilities. One of my major goals is to develop a typing error detection and correction system for Mandarin Chinese text. An intermediate step is to develop a simplified-traditional text conversion system. Since there is one-to-many mapping when converting a text from simplified to traditional character set, the ambiguation needs to be resolved. This problem is extremely similar to the typing error detection and correction problem, but much smaller in scale and much simpler in complexicty.

Now back to the research. I have chosen word segmentation in Chinese as my main research theme with care. Words are out there and you need to identify them, but there are no word boundaries. How do you identify words without word boundaries marked in the script? This is a perfectly valid theoretical issue and many interesting factors come into play: serial vs. parallel, top-down vs. bottom-up, linguistic vs. non-linguistic , morphological vs. syntactic, and perceptual vs. semantics processing. Chinese word segmentation is not only important in Chinese, but also a good test case for script-independent general psychological theories.

Research and software development support each other in this way. Algorithms discovered in research will be used directly implemented in software, if possible. If direct implementation is unfeasible, human algorithms can still serve as cues in developing practical, efficient algorithms. On the other hand, the techniques used in software development are quite similar to those used in implementing psychological algorithms. There is much positive, bidirectional transfer.

Where do we go from here? This isn't where we intended to be ... (Excerpt from Evita lyrics: You Must Love Me). This isn't where I intended to be, either, but I am already here. So where do I go from here? It will be another story in the form of a two-year plan, but I have already said some.