Chia-hao Shen
Focus on Natural Language Processing, Automatic Speech Recognition
Data Scientist/Machine Learning Engineer
New York, USA
[email protected]
(+1) 917-769-5181
Experience
Data Scientist, Nov. 2017- April. 2018
Yoctol Info
- Working in a team with 6 people to optimize the intent classification and entity extraction model used for general/customized chatbots in Yoctol.
- In charge of the structuring and building of the server.
- Working on research topics and production development with Python, Tensorflow, C++.
GPU Cluster Manager, 2015 - 2017
NTU Speech Lab
- Built and managed the GPU cluster with another administrator.
- Completion of the connection between GPU clusters. Maintained whole updating scripts.
- Skill used: Python, bash, javascript, perl
Teaching Assistant, Sep. 2016-Feb. 2017
Machine Learning Course in NTU
- Designed and managed homework with topic semi-supervised image classification with CIFAR10.
- Set up baseline with performance up to 80% accuracy using ImageNet structure.
- Skill used: Python, Tensorflow, numpy
Teaching Assistant, Feb. 2016 -Jul. 2016 Deep Learning Course in NTU
- Designed and managed homework with topic RNN-based TIMIT phoneme sequence prediction.
- Set up baseline with performance up to 84% phoneme sequence accuracy.
- Skill used: Kaldi, numpy, theano
EDUCATION
National Taiwan University, 2015-2017
Master Science of Electrical Engineering
- GPA: 4.1 / 4.3
- Graduate Thesis: Audio Word2Vec: Unsupervised Learning of Audio Segment Representations using Sequence-to-sequence Autoencoder
- Relevant Courses: Machine Learning, Machine Learning and Having it Deep and Structured, Natural Language Processing
National Taiwan University, 2009-2013
- GPA: 3.8 / 4.3
- Relevant Courses: Introduction to Digital Speech Processing, Data Structure and Algorithms, System Programming
SKILL
Programming
Python, C/C++, Java, Bash
Deep Learning Toolkit
Tensorflow, Keras, Kaldi, Sklearn
Language
Traditional Chinese
English
Publications
- Phonetic-and-Semantic Embedding of Spoken Words with Applications in Spoken Content Retrieval,
SLT 2018
- Language Transfer of Audio Word2Vec: Learning Audio Segment Representations without Target
Language Data, ICASSP 2018
- Audio Word2Vec: Unsupervised Learning of Audio Segment Representations using Sequence-tosequence
Autoencoder, Interspeech 2016
