Welcome to Tao Chen's Homepage

I am working at Google Research, and focusing on Natural Language Processing and Information Retrieval.

I was a postdoc at the Center for Language and Speech Processing (CLSP), Johns Hopkins University (JHU), working with Prof Mark Dredze. Prior to that, I obtained my Ph.D from School of Computing, National University of Singapore (NUS) in April 2016. I was advised by Prof Min-Yen Kan and was affiliated with the Web, Information Retrieval / Natural Language Processing Group (WING).

My research spans the fields of social media, multimedia, natural language processing, health informatics and applied machine learning. My Ph.D thesis is about analyzing, classifying, modeling and recommending image tweets. During my postdoc in JHU, I was working on social media analysis for public health and natural language processing for healthcare.

Here is my CV and Google Scholar profile.

Image tweet, the tweet that embeds an image, is a popular form of communication in microblog platforms. I conduct comparative studies on image tweets from Twitter and Chinese Weibo [Ph.D Thesis], categorize the correlation of the image and its accompanying text [MM'13], model the generative process of image tweets [AAAI'15], and recommend interesting image tweets to users [MM'16] [MM'17].

What makes a tweet worth sharing? We study the content of tweets to uncover linguistic tendencies of shared microblog posts (retweets). We build a classifier to categorize tweets into four pre-defined types and build a model to predict the retweetability of a tweet [LSM'12].

We study how users of multiple online social networks (OSNs) employ and share information by studying a common user pool that use 6 OSNs – Flickr, Google+, Instagram, Tumblr, Twitter, and YouTube. We explore and analyse the factors that affect user behaviour. In particular, we take a temporal and topical approach in studying the sharing behaviour of users, and we show how users exhibit distinct patterns on different networks [ASOMAN'15].

We explore the rich evidence source of aspects in user reviews to improve top-N recommendation. By extracting aspects from textual reviews, we enrich the user–item binary relation to a user–item–aspect ternary relation. We model the ternary relation as a heterogeneous tripartite graph, casting the recommendation task as one of vertex ranking. We devise a generic algorithm for ranking on tripartite graphs — TriRank — and specialize it for personalized recommendation [CIKM'15 ] .

We propose WordNews a web browser extension that allows readers to learn a second language vocabulary while reading news online. Injected tooltips allow readers to look up selected vocabulary and give interactive tests to assess vocabulary mastery. In particular, we propose techniques for word sense disambigution (WSD) to aid transation quality and constructing appropriate distractors for multiple-choice word mastery tests [NLP-TEA'15] [NLP-TEA'16].

Our WordNews project was funded by VaSCo (Validating Startup Concept) award (S$10,000) from National University of Singapore!

I was engaged in a live project that collects SMS Messages for a public research corpus. As of May 2015 , we have collected 45,718 English SMS and 31,465 Chinese SMS. To the best of our knowledge, it is the largest English and Chinese SMS corpus in the public domain [LRE'13].

Check our GitHub page to download the dataset.

A Java program that parses dblp.xml and stores inproceedings into mysql database.

In early 2010, I implemented this program for my undergraduate thesis, and wrote a post in my Chinese techinical blog. Since then, I received a lot of emails, requesting for the code. Of course, I emailed them the code:)

To make this program benifit more people, especially English speakers who will never be able to search my Chinese blog, I released the code in Google Code and recently transferred to Github. Meanwhile, I am happy to receive comments, suggestions, and bug reports from the world.

An android app that backs up user's SMS messages from phone to his/her Gmail.

While this app was originally implemented to faciliate users to contribute SMS messages to our NUS SMS Corpus, it functions purely as a backup software, sending SMS to user's Gmail as draft emails.

No worries about the privacy issue. On one hand, the app automatically encrypts sensitive information (e.g., emails, phone numbers) in messages . On the other hand, the app does not send the messages directly. We leave the decision to users to send the draft email to us or not.

I like sports. Previously I played table tennis and badminton regularly. More recently I spent more time in the gym than playing ball games.

I love travelling and have been to over 10 countries. When I am in a new place, I like viewing the landscape and people by my eye and camera， as well as visiting museums to know the history, culture and art in a documental way.

Here are my conference and internship debrief given in group meeting.

Conference Ranking, National University of Singapore
Journal Ranking, National University of Singapore
NLP Conference Calendar

Welcome to my research homepage!

Image Tweets Analysis, Categorization and Modeling

Retweets Analysis and Prediction

User Behaviour Analysis across Multiple Social Networks

Recommendation

Language Learning

NUS SMS Corpus

DBLP XML Parser

SMS Collection for Corpus

Personal

Slides

Links