My name is Ming-Wei Chang. I am currently a Research Scientist at Google Deepmind , working on fun research problems related to machine learning and artificial intelligence!

Twitter

Google Scholar

Semantic Scholar

News

Selected Projects (Full List)

Here are some fun projects I have worked on with many awesome researchers and students.

OVEN and InfoSeek. project1 project2

Can you pretrained models understand textual and visual knowledge at the same time?

SUTI paper

Dreambooth with in-context learning (no fine-tuning required!)

Pix2Struct. paper

It is possible to parse an website from pixels only.

Promptagator. paper

Propose the first task-specific prompt for retrieval. Eight examples are enough for buidling a pretty good retriever!

FRUIT paper

FRUIT is a new task about updating text information in Wikipedia. A really fun project! (NAACL 2022 best new task paper)

BERT. paper

BERT is a framework for pre-training deep bidirectional representations from unlabeled text. BERT achieves state-of-the-art results for 11 nlp tasks when it was published. (NAACL 2019 best paper)

Zero-shot entity linking. paper

The power of text understanding makes zero-shot entity linking finally possible. (ACL 2019 best paper candidate)

Semantic parsing for knownledge base. paper

By applying an advanced entity linking system and a deep convolutional neural network model, this semantic parsing system outperformed previous methods substantially when it was published. (ACL 2015 outstanding paper)

Semantic parsing using weak supervision. paper

This project shows that learning with a weak feedback signal is capable of producing strong semantic parsers. This was very surprisingly to me at that time.

Load forecasting using SVM. paper

My first (or second?) research project (in 2001!) under the supervision of the amazing Chih-Jen Lin. In this project, we use SVM to predict the power needed to balance the supply and load for powerplants. Winner of the EUNITE competition 2001.