Xing

Xing

Ph.D. Candidate at IU School of Informatics and Computing

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

Python basics

5 minute read

Published:

Author: Xing Yu

How to setup I101 appliance

less than 1 minute read

Published:

I101 is a virtual machine built on Linux to help students in I101 learn coding, database, and creating websites on LAMP structure. This tutorial helps you setup the appliance.

Introduction to python

4 minute read

Published:

Python (version 3 or later) is the main programming language that we will learn and use in I101. Python is a interpretive language that doesn’t need to be compiled before running. Its grammar is relatively intuitive. And it is a very popular programming language used in informatics and computer science. This tutorial is a start point for you to get familar with programming in general.

Blog Post 4

less than 1 minute read

Published:

### #— title: ‘Blog Post number 4’ date: 2015-08-14 permalink: /posts/2012/08/blog-post-4/ tags:

  • cool posts
  • category1
  • category2 #—

portfolio

publications

Using data from social media websites to inspire the design of assistive technology

Published in Proceedings of the 13th Web for All Conference, 2016

The rapid accumulation of user generated content on the Internet provides researchers with abundant information to extract knowledge from. It also provides HCI and accessibility practitioners with a new direction to explore and understand users requirements beyond traditional approaches. In my work, I create a tool that consists of both text-mining and machining learning methods to extract essential focuses of design out of the data collected from social networks. This tool can be used at the initial stage of a product design life-cycle by designers to collect key design aspects at a fairly low cost.

Download here

Comparing Community-based Information Adoption and Diffusion Across Different Microblogging Sites

Published in Proceedings of the 27th ACM Conference on Hypertext and Social Media., 2016

The proliferation of social media is bringing about significant changes in how people make sense of their world and adopt new information. However, social, cultural and political divisions continue to separate users and information into different social media systems. Twitter and Facebook, for example, are strictly forbidden in mainland China. As a result, 21.97% of all world-wide Internet users are thus excluded from participating in these platforms. In this study, we investigate whether the dynamics of information diffusion, modeled as the adoption patterns of topical hashtags, differ between the communities of the mentioned social media sites as a result of this separation. Specifically, we compare Weibo and Twitter, the two largest micro-blogging sites serving respectively the Chinese population and the rest of the world, by exploring the similarities and differences of how their respective users adopt new information. By leveraging sophisticated community detection algorithms and heterogeneous graph mining methods, we investigate and compare how the different characteristics of these communities influence information diffusion and adoption. Experimental results show that while community-specific information influences topic diffusion and adoption in both environments, novel features, extracted from heterogeneous graph based communities, have a greater effect on Weibo information adoption than Twitter. We also find that users sharing hashtags is an important factor in information diffusion on both Twitter and Weibo, whereas user mentions are important for Weibo, but less so for Twitter. Overall, we conclude that Weibo and Twitter differ sharply in how their users adopt information in response to similar factors.

Download here

Exploring Customers’ Search Behavior on a Large E-Commerce Website

Published in Proceedings of the 22nd Americas Conference on Information Systems, 2016

In this study, we have teamed up with a large e-commerce website, Walmart.com, and analyzed one-day search logs in 2014. Different measures were compared among three different types of devices—desktops, tablets, and mobile phones. The discrepancies as well as similarities regarding search time, search location, click behaviors, and query behaviors are reported. The implications of the findings and the future work are discussed at the end.

Download here

Predictive Analytics of E-Commerce Search Behavior for Conversion.

Published in Proceedings of the 22nd Americas Conference on Information Systems., 2017

This study explores online customer search behavior on a large e-commerce website—Walmart.com. In order to more accurately predict customer purchase conversion based on their search behavior, we adopt a modern machine-learning technique, random forest, as well as logistic regression to develop two computational models. We also integrate information retrieval literature to propose metrics to quantify online consumers’ search behavior. Results show that the random forest model performs better with a very high accuracy rate (76%) in predicting customers who will purchase the item they browsed. Among all the predictors, page and session dwell time, user type, click entropy, and click position are the strongest influential factors for the conversion behavior. The findings suggest that, with the enhanced metrics and modeling approaches, search behavior could offer strong cues about customers’ purchasing decision. Additionally, the findings also suggest operational implications about how to accommodate and induce the desired search behavior with the e-commerce website.

Download here

Internal/External information access and information diffusion in social media

Published in iConference 2017, 2017

As social media platform not only provide infrastructure but also actively perform algorithmic curation for profit and user experience, it leads to an information filter bubble phenomenon: users are trapped in their own personalized bubble and are exposed only to the opinions that conform their beliefs and interests, thus potentially creating social polarization and information islands. However, filter bubbles hardly restrict all the users in a large social network, some information explorers can break the bubble and bring external global knowledge back to the internal network. In this paper, we investigate this assumption via hashtag adoption prediction. First, we construct a heterogeneous graph and extract 17 features to describe the event of hashtag adoption. Then, we generate learning instances and train a lasso regression model to do prediction. Preliminary results show that information explorers are more likely to adopt new hashtags than others, thereby more internal and external information can be diffused via these special users.

Download here

Designing Leaderboards for Gamification: Perceived Differences Based on User Ranking, Application Domain, and Personality Traits

Published in 2017 CHI Conference on Human Factors in Computing Systems, 2017

Leaderboards, a common gamification technique, are used to enhance engagement through social comparisons. Prior research has demonstrated the overall utility of leaderboards but has not examined their effectiveness when individuals are ranked at particular levels or when the technique is applied in different application domains, such as social networking, fitness, or productivity. In this paper, we present a survey study investigating how preferences for leaderboards change based on individual differences (personality traits), ranking, social scoping, and application domains. Our results show that a respondent’s position on the leaderboard had important effects on their perception of the leaderboard and the surrounding app, and that participants rated leaderboards most favorably in fitness apps and least favorably in social networking contexts. More extraverted people reported more positive experiences with leaderboards despite their ranking or the application domain. We present design implications for creating leaderboards targeted at different domains and for different audiences.

Download here

Understanding and Classifying Online Amputee Users on Reddit

Published in 2017 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining 2017 (ASONAM '17), 2017

Accessibility researchers have difficulty recruiting representative participants with disabilities given their scarcity. The rich information on social media provides accessibility researchers with a new approach to collecting data about these populations. Because social media is used by multiple stakeholders, a major barrier to this approach is differentiating representative users who have disabilities from unrepresentative users who do not. We (1) introduce an empirical study that compares representative users who are amputees with unrepresentative users in terms of linguistic behavior, online interaction, and community characteristics on Reddit and (2) develop a feature extraction method based on statistical analyses and graph mining to classify representative users. Those features allow us to detect amputees using a supervised learning method with an overall accuracy of 88% in amputee-related subreddits. Our findings improve our understanding of anonymous online users with physical disabilities, and can inform better tools for online data collection for accessibility researchers.

Download here

talks

ASONAM 2017

Published:

Presentation at ASONAM 2017.

teaching