Projects

OpentoWork Web App
A job recommendation web app that matches job postings with user profiles based on skill sets, visa sponsorship, and other attributes using LLM models.
GitHub Repo

YouTube Trends Dashboard
A visualization showing what categories trend, variation of trends over the years, and trending patterns across countries.
Dashboard

Pest Detection
This project employs machine learning and perception methods to improve early detection of insect pests from imagery. The research investigates a problem in computer vision where models trained on image datasets with "perfect" conditions may not perform well in the wild due to deformations and occlusion. Another problem we examine is how to deal with limited insect data. The solution proposed is to use synthesized images in the training set to represent arbitrary viewpoints and deformations, thus narrowing the gap between in-sample and out-of-sample errors.
Poster

Enabling Crop Analytics at Scale

Cropland mapping
This project aims to help improve agricultural extension services for African smallholder farmers while providing a scalable, sustainable model for generating open agricultural maps and ground-truth data.
I help improve the accuracy of cropland mapping ML model with Prof Lyndon Estes.
AIRG Research Group Cropland Mapping Mapping Africa Example GitHub repo

BookEz

Book-EZ
A scalable online bookstore web application featuring database management to improve organization and data processing as measured by an increase in productivity by 20%.
Technologies: HTML, CSS, Bootstrap, JavaScript, jQuery, NodeJS, MongoDB.
GitHub repo

YouTubeCommentAnalysis

YouTubeCommentAnalysis Python Package
A Python package to decrease time processing YouTube’s comments data as featured in CHI publications by doing text cleaning, emotion analysis, and topic modeling.
Technologies and libraries: YouTube Data API, nltk, gensim, pandas
PyPI page GitHub repo

Investigating Multimodal Interactions and Parasocial Attractiveness in YouTube ASMR Videos
This work annotates and analyzes the interaction modalities and parasocial attractions of 2663 ASMR YouTube videos to characterize viewers' experiences. YouTube comment sections are analyzed to compare viewers' responses to different ASMR interactions. The result demonstrates 3 experiences that ASMR videos can provide as published in CHI Conference Paper.
CHI22 Paper CSCW2021 Poster

#StayHome #WithMe: How Do YouTubers Create Videos for COVID-19 Loneliness?
Loneliness threatens public mental well-being during COVID-19 isolation. YouTube creators responded to the call of #StayHome #WithMe movement and made myriad videos for people who feel lonely or bored at home. Grounded on Weiss's loneliness theory, this work analyzed 1488 SHWM videos on YouTube to examine video-sharing as a pathway to social provisions. Findings summarize popular types of videos to engage viewers suggested for supporting viewers’ mental well-being in disasters as measured by CHI Conference Paper.
CHI21 Paper Clark Winter Fest 21 Poster

Work Experience

Research Assistant
Clark University Graduate School of Geography | Worcester, MA
  • Improving machine learning prediction accuracy by implementing image segmentation and analyzing data from PlanetScope and Google Earth Engine API.
  • Optimizing image processing pipeline runtime by 70% using Python libraries.
Undergraduate Researcher
Clark University Department of Computer Science | Worcester, MA
  • Designing scalable systems to improve object detection in the agricultural domain by implementing GAN and R-CNN using the TensorFlow library.
  • Sponsored by Clark University LEEP Fellowship.
Business Analyst Intern
NexLab Technology Software Team | HCMC, Vietnam
  • Improved software product quality by mapping business functions to software features.
  • Reduced debugging time by 5% by performing quality tests and communicating with the design and development teams.
Undergraduate Researcher
Clark University Department of Computer Science | Worcester, MA
  • Analyzed YouTube video creation characteristics using grounded theory to improve media productions as featured in CHI Conference Papers.
  • Wrote text classification interface to reduce time creating labels for Amazon MTurk by 30% in JavaScript.
  • Examined viewers’ engagement by analyzing 8M YouTube comments from over 10,000 videos using Python libraries.
Academic Technology Assistant
Clark University Office of Information Technology Services | Worcester, MA
  • Maintained quality control of 3 teaching websites.
  • Troubleshoot problems coming from students, professors, and school staff.

Publications and Posters

Nguyen B. Ha, Philippe Nziza, 2023. Experiments in Synthesized views for Improved Pest Detection. In ClarkFEST 20223, April 26, 2023. Worcester, MA, US. [poster]

Shuo Niu, Ava Bartolome, Nguyen Binh Ha, Keegan Veazey. 2022. Close-up and Whispering: An Understanding of Multimodal and Parasocial Interactions in YouTube ASMR videos. In CHI ’22: ACM CHI Conference on Human Factors in Computing Systems, April 30 - May 6, 2022, New Orlean. ACM, New York, NY, USA. [paper]

Ava Bartolome, Nguyen Binh Ha, Shuo Niu, 2021. Investigating Multimodal Interactions and Parasocial Attractiveness in YouTube ASMR Videos. In CSCW ’21: ACM Conference on Computer-Supported Cooperative Work & Social Computing, October 23 - 27, 2021, Online virtual. [paper] [video]

Shuo Niu, Ava Bartolome, Cat Mai, Nguyen Binh Ha. 2020. #StayHome #WithMe: How Do YouTubers Create Videos for COVID-19 Loneliness? In CHI ’21: ACM CHI Conference on Human Factors in Computing Systems, May 08–13, 2021, Online virtual. ACM, New York, NY, USA. [paper] [video]

Ava Bartolome, Cat Mai, Nguyen Binh Ha, 2021. #StayHome #WithMe Investigating How YouTubers Help Deal with COVID-19 Loneliness. In Clark University Winter Fest 2021, February 5, 2021. Online virtual. [poster]