Zhuyun Dai

Language Technologies Institute, Carnegie Mellon University

Home
Publications
Experiences
Projects

 

PhD Software Intern, 2018, Facebook

Summer Intern at Facebook Search Team, Seattle

 

Research Assistant, 2014, Peking University

An Item-based Joint Aspect/Sentiment Model for Online Reviews

With the rapid development of platforms for Electronic Commerce and on-line Ratings and Reviews, an overwhelming amount of user-generated reviews have been posted to the Web. People comment on a variety of aspects with different sentiments for each aspects. Information about those aspects and sentiments show details about the item being reviewed. However, most of the online reviews are plain text without any structure.

In this research we propose an Item-based Joint Aspects-Sentiment (IAS) Model based on Latent Dirichlet Distribution. IAS model automatically discovers aspects and their corresponding sentiments. It is an un-supervised model and does not require manually labeled data. We evaluated IAS model on reviews from two domains: hotels and restaurants. Experiment results have shown that IAS model works well in sentiment classification, aspect detection and aspect-specific sentiments words discovery.

 

PM Intern, 2014, Microsofts

Ratings & Review API

As a Program Manager (PM) Intern, I designed a suite of Ratings & Review API to support Microsoft Cooperators. I also drove the development and test for the API.

 

Research Assistant, 2014, Peking University

A Management Model for SDN-based Data Center Network

Current data centers often employ layered SDNs (Software-Defined Networks). However, information overwhelming has been a limitation to SDN deployment. This resarch focused on designing a management model for data center networks. In this model, regional networks on lower layers will be aggregated and viewed as single switches to upper layers. Management information will be divided into three parts, which can be seen by network managers, regional controllers and tenants, respectively.

 

Student Intern, 2013, Carnegie Mellon University

A Cognitive Assistive System for Monitoring the Use of Asthma Inhaler

Many patients do not use the asthma inhaler correctly. Incorrect using of inhalers significantly weakens the effect of the drug.Our goal is to design a system which monitors and gives advice to inhaler users. We used RGB and depth cameras of Kinects to collect video data. Audio data were also recorded through a microphone. Through signal processing and pattern recognition methods, we built a system that is able to guide users and recognize their errors. The experimental results show that our system works accurately and efficiently in offline situations, while a system for online situations remains to be developed.