Jiarui Gao
|
Hi there! My Chinese name is 高嘉蕊, which means beautiful blooming flowers. I grew up in Changchun City, a lovely snowy place in northeast China. I'm a software engineer at Google. I'm passionate about connecting people with techonology, and building products that are equally accessible to everyone. I obtained my master degree in software engineering at Carnegie Mellon University, Silicon Valley. I received my bachelor's degree in computer science from Fudan University, Shanghai. I was advised by Prof. Yanwei Fu and Prof. Yu-Gang Jiang to work on machine learning and multimedia then. CV / LinkedIn / GitHub |
|
The profile picture is taken by my sweet boyfriend Xingyi Zhou at Santa Monica.
Last updated 12/31/2020 by Jiarui Gao
Google LLC, Sunnyvale, CA
Developed a manual review tool for Engagement Rewards By Google to visualize per user abuse information.
Established fully tested back-end RPC services to retrieve and process data from Spanner database.
Integrated with two frameworks to securely access user data and utilize Google-wide abuse protections.
Morgan Stanley, Shanghai
Developed a Java web application for Outside Business Interest system as an effective PoC.
Proposed a novel graph algorithm in AngularJS2.0 on front-end to render decision processes with a config- urable set of tree-like rules.
Designed a new database schema using DB2 temporary tables on back-end to make history data traceable.
In this paper, we propose a new architecture–Frame-Transformer Emotion Classi cation Network (FT-EC-net) to solve three highly correlated emotion analysis tasks: emotion recognition, emotion attribution and emotion-oriented summarization. We also contribute a new dataset for emotion attribution task by annotating the ground-truth labels of attribution segments.
We propose a new neural approach Frame-Bi-stream Emotion Attribution-Classification Network(BEAC-Net), an end-to-end trainable neural architecture that tackles emotion attribution and classification simultaneously with significant performance improvements. Also we propose an efficient dynamic programming method for video summarization based on the output of A-Net. To establish a good benchmark for emotion attribution, we re-annotate the Ekman-6 dataset with the most emotion-oriented segments which can be used as the ground-truth for the emotion attribution task.