While every bit in real world matters...

Currently, I am a first year master student in Computer Science at University of Southern California. Previously, I am lucky to have opportunities to work with Dr. Wei-Yun Ma as a research assisstant at Chinese Knowledge and Information Processing Lab (CKIP Lab), Academia Sinica. I received my Bechalor's Degree at National Tsing Hua University. See my resume for further information.

Recent Works

CosmEL: Cosmetics Entity Linking

Designed a distant supervision model applying noisy labeling using PyTorch. This work mainly focus on cosmetics domain and most cosmetic entities are offered from PIXNET StyleMe.

Wikipedia Title Embeddings

Proposed an algorithm to improve word embedding on presenting Wikipedia titles in Python and also improved spearman correlation score on word similarity task by 10%.
The Chinese Wikipedia Title Embeddings are avalible at [link].

Chinese Word Embeddings Benchmarks Creation

Created large evaluation benchmarks for Chinese word embeddings in both word similariy task and analogical reasoning task by manually translating from English benchmarks.
The benchmarks are available at [link].

Answering Machine

Designed a simulated intelligent program able to answer simple WH-questions and make sentences with given words in Java which can achieved around 90% of correction rate. Also, built the knowledge base for this program by SQLite.

Traffic Alarm

Developed a real-time traffic signal detection algorithm for drivers as an Andriod app. In order to achieve real-time performance on Android devices, system is optimised to cope with 20 fps.

Face Recognition

Designed a gamification program that can categorise people's photos in Java. Also, constracted an interface that can easily demenstrate each person's photos from database.
The demo vedio is available at [demo].


Word Embedding Evaluation Datasets and Wikipedia Title Embedding for Chinese

Chi-Yen Chen and Wei-Yun Ma

LREC 2018 [paper]

Embedding Wikipedia Title Based on Its Wikipedia Text and Categories

Chi-Yen Chen and Wei-Yun Ma

IALP 2017 [paper]
Best Paper Award