Thesis Topic: Unsupervised Learning and Modeling of Knowledge and Intent for Spoken Dialogue Systems

Spoken dialogue systems typically use predefined semantic concepts and their structures to parse users' natural language inputs into unified semantic representations, where domain experts and professional annotators are often involved, resulting in higher cost of the system development. This work focuses on answering the following questions:

  1. Given a collection of unlabelled raw audio, can we use the available knowledge resources to automatically induce the semantic concepts and learn the structures?
  2. Then how can we use the learned information to develop spoken language understanding models for spoken dialogue systems and to improve human-machine conversations?
The thesis mainly focuses on four important stages: semantic concept induction, structure learning, semantic decoding, and behavior modeling, where concept induction and structure learning define the domain-specific organized knowledge, and with the learned knowledge, semantic decoding and behavior prediction enable the developed systems to understand users' spoken language inputs and further provide better interactions.

About Me

I hold two M.S. degrees and a B.S. degree, and recently my boyfriend, Che-An (JAST) Lu, became my fiance !

M.S. from Carnegie Mellon University

I earned a M.S. in Language Technologies from School of Computer Science in 2013. The photo was taken near the fence in the center of campus, and the word in the fence is the traditional chinese characters about "Taiwan".

M.S. from National Taiwan University

I earned a M.S. in Computer Science and Information Engineering in 2011. The photo is my lovely family, my younger brother with his bachelor degree and my parents.

B.S. from National Taiwan University

I earned a B.S. in Computer Science and Information Engineering in 2009. The photo includes my classmates from the same department, taken in front of the famous clock in our university.