Chieh-Yang Huang
Research Assistant
Affiliation:
Insitute of Information Science,
Academia Sinica
Email:
Research Interests:
Natural Language Processing, Deep Learning, Computer Assisted Language Learning, Web Programming
Hello! I'm Chieh-Yang
I'm a research assistant in Insititute of Information Science, Academia Sinica, Taiwan. As a data scientist, my research interests are Natural Language Processing (NLP), Deep Learning and Computer Assisted Language Learning (CALL). Also, I am passionate about designing and building websites and web applications, which leads me into a web developer.
EXPERIENCE
EDUCATION
2015-now
Research Assistant
Insitute of Information Science,
Academia Sinica
I'm a research assistant in Natural Language Processing Lab. I am working on GiveMeExample and EmotionPush project now.
2012-2014
Research Student
NTU Speech Processing Lab
I was responsible for a CALL project, which helped learner improve poorly-pronounced syllables.
2013-2014
Research Student
NTU Communication & Multimedia Lab
I worked on two different projects in two semester, which were "light field image multi-label assignment" and "shape-preserving as-projective-as-possible image stitching".
2014-2014
Research Student
NTU Multimedia Information Retrieval Lab
I was responsible for the project "music separation", which wanted to separate the human voice from a song.
2010-2014
Bachelor in Electrical Engineering
National Taiwan University
I graduated from NTUEE (Department of Electrical Engineering, National Taiwan University), the best EE school in Taiwan.
2007-2010
High School
Jianguo High School
Jianguo High Shool is the best boy high school in Taiwan.
SKILLS
Programming Language:
Python, Java, C/C++
Language:
Chinese, Taiwanese, English
Web Programming:
HTML, JavaScript, CSS
Other:
PhotoShop, Illustrator, Unity
PROJECT
Current Research Project
GiveMeExample:
Automatically Suggesting Example Sentences of Near-Synonyms for Language Learners
GiveMeExample is proposed to automatically recommend example sentences for language learners in order to help them understand the difference among near-synonyms.
System URL: http://givemeexample.com/GiveMeExample
EmotionPush:
Automatically Providing Color-Based Emotion Cues for Messaging Applications
We introduce EmotionPush, a machine-learning-powered system that automatically conveys users’ emotions in messages by color-based emotion cues.
System URL: Please Click Here
Student Research Project
Light Field Image Multi-Label Assignment
using Graph Cut
Advisor: Prof. Yung-Yu Chuang
The purpose of this project is to extend the original Graph Cut algorithm to light field image. Light field image contains depth information which is very helpful for separating objects which have the similar color. In this project, we try to use depth information by considering the property of light field image instead of calculating depth directly.
Selected Courses
Acceleration of Context-Aware Saliency Detection Using PatchMatch
Course: Advanced Digital Signal Process
Saliency Map reveals the important object in a figure. In the context-aware approach, it costs lots of time to calculate the saliency map. As a result, I proposed a method to accelerate this approach by PatchMatch algorithm.
< More >
Artificial Intelligence for Draughts
Course: Artificial Intelligence
Draughts is a strategy board games for two players. The rules of draughts are quite complex which make its AI hard to implement. In this project, we implement three different algorithms, random agent, minimax agent and q-learning agent.
< More >
Image Completion with PatchMatch
Course: Digital Image Processing
PatchMatch is proposed by Adobe, which is used to context-aware image editing. In this project, we try to re-implement PatchMatch algorithm and use it to reconstruct and fill in the blank part of the original image.
< More >
Shape-Preserving As-Projective-As-Possible Image Stitching
Advisor: Prof. Yung-Yu Chuang
The most famous two image stitching algorithms are AutoStitch and APAP. However, AutoStitch fails to maintain the line-preserving property and APAP distorts the image seriously. As a result, in this project, we introduce a half-projective wrapping algorithm to find the balance between several properties.
A Dialogue Game Framework with Personalized Training Using Reinforcement Learning for Computer-Assisted Language Learning
Advisor: Prof. Lin-Shan Lee
This system proposed an improvement mechanism for poorly-pronounced syllables on a dialogue game framework.
