Sung-Young Jung

Affiliation

  • Ph.D student in the Intelligent Systems Program at University of Pittsburgh
  • Research Assistant in the School of Computing, Informatics, and Decision Systems Engineering in Arizona State University
  • My adviser is Kurt VanLehn

  • e-mail: (chopin at cs dot pitt dot edu) or (syjung5 at asu dot edu)
  • tel: +1-480-712-0154
  • Vita

    Research Interest

     My research interest covers several areas in artificial intelligence such as knowledge representation, natural language understanding, knowledge acquisition, teachable agents, machine learning, and intelligent tutoring systems. I am especially interested in AI systems performing tasks using knowledge acquired from users. My Ph.D. research is to find a way to use natural language to represent knowledge in an ITS (intelligent tutoring system) so that authors can add knowledge easily, and the system can read and solve physics problems using the acquired knowledge (Ph.D Thesis, defense). In this research, all the knowledge in the system was written in natural language to overcome the knowledge acquisition bottleneck, which is a unique contribution to the AI area. I am trying to extend this idea further for a general solution to knowledge acquisition and authoring problems; using natural language to represent not only factual knowledge but also procedural knowledge will enable users to teach procedural tasks to teachable agents.
  • Using natural language as knowledge representation in an ITS (Ph.D Thesis, defense(pptx, ppt))
    This thesis presents an approach to represent knowledge in unconstrained natural language so that an ITS can solve physics problems using knowledge acquired from authors. The system can read physics problem statements, generate implied facts, and solve the problems. All the knowledge in the system including principles was written in natural language to overcome the knowledge acquisition bottleneck. Total 92 physics problems from textbooks were solved correctly and some of them were solved automatically without adding new knowledge.
  • Programming in unconstrained natural language: Natural-p This project is to extend the idea of knowledge representation in natural language to procedural knowledge. In this approach, the system can read natural language text and execute it generating the same output when a human reads it. A user can add required background knowledge in natural language in a way teaching the system, and a user can teach the system not only what to do but also how to do. This technology is important for teachable agents. Natural-p was driven from my thesis project, and it was started as a personal project.
  • Machine Learning: Preference Modeling (IEEE TKDE, IEEE ICDM2002, Presentation) Previous preference measures had a problem that the preference values cannot be interpreted intuitively. The presented statistical model provides an intuitive preference measure not only for positive but also negative preference. TV audience data was analyzed to measure audience preference on each TV program based on synopsis, and both word preference and text preference were measured. The model was tested to predict audience's TV program selection.
  • Computational Linguistics: Transliteration (COLING2000), MRF POS Tagging (COLING1996)
  • Artificial Life: Evolutionary Artificial Brains (Artificial Life, MIT Press)

    • Publication

    My Piano Playing

    Chopin Etude Op. 25 No. 12 (Ocean)
    Chopin Nocturn No. 2
    Chopin Etude Op. 10 No. 12 (Revolutionary)     .

    .