DC Housing Policy Chatbot

A document QA chatbot demo built with LangChain and ChromaDB for analyzing DC housing policy documents.

Policy Change Index for North Korea (PCI-NKO)

PCI-NKO uses deep learning models and AI to analyze over 80,000 articles from North Korea's state newspaper. The project detects subtle changes in policy emphasis, providing unique insights into the opaque regime's future actions and priorities.

Harnessing AI to Map Political Ideologies

This project harnesses AI to estimate U.S. House members' ideological preferences. By analyzing public statements and employing pairwise comparisons, the study showcases Large Language Models' potential for accurate and efficient analysis in traditionally qualitative political research.

Comprehensive Neural Network Approaches for Fake News Detection

This project compares 12 neural network models for fake news detection, from simple CNN/RNN to advanced Transformers. It focuses on improving performance on small, topic-specific datasets like FA-KES, using techniques such as data augmentation and transfer learning to enhance model generalization.

Decoding Political Issue Framing in News Media using Word Embeddings

This project employs word embedding techniques to analyze how news media frame political issues. By training models on articles from 2000-2019, it quantifies media bias through sentiment associations with political terms, exploring how outlets attach different sentiments to the same issues.

Measuring Political Sophistication through Textual Complexity

This project replicates Benoit et al. (2019) and reproduces a model that quantifies the sophistication of political texts using pairwise comparisons and a Bradley-Terry statistical model. The replication validates the original findings while identifying potential improvements.

Tracking the Changing Meanings of Political Concepts with Word Vectors

This replication study implements four time-aware word2vec models to analyze the evolving meaning of "equality" in historical newspapers. We updated the original code and confirmed the original findings while reducing computation time from 46 to 12 hours.

Tutorial on stLDA-C Model for Congressional Tweet Analysis

This tutorial introduces the stLDA-C model for analyzing short texts, demonstrating its application to tweets from U.S. Congress members. It guides learners through implementing this advanced topic modeling technique on a dataset of tweets, showcasing user clustering and topic estimation.

Walk the Route: Visualizing Chinese Migration at the U.S. Southern Border

This project analyzes and visualizes the recent surge of Chinese migrants at the U.S. southern border using Google Trends, migrant interviews, and social media data. It combines advanced data collection, processing, and visualization techniques to provide insights on this emerging trend.

Evaluating the Effectiveness of Washington D.C.'s Safe Passage Program

This ongoing study preliminarily assesses D.C.'s Safe Passage program, aimed at protecting students from gun violence during school commutes. Using difference-in-differences, temporal, and spatial analyses, it examines the program's impact on student retention and community safety perceptions.

Determinants of Childbearing Intentions Among Married Working Women in South Korea

This study replicates and extends research on childbearing intentions of married working women in South Korea using logistic regression and decision tree methods. It analyzes data from the Korean Longitudinal Survey of Women and Families to identify key factors influencing fertility intentions.

The Impact of Social Media Use on Students' Academic Performance: A Cross-Country Analysis Using PISA 2018 Data

This study examines the effects of social media usage on students' academic performance across 50 countries using PISA 2018 data. Employing propensity score matching, it reveals significant cross-country variations in how social media use for school-related communication impacts performance.

Analyzing the Effect of Funding on School Performance in DC Public Schools

This team project examines the relationship between per-pupil expenditure and school performance, as measured by STAR scores, in Washington DC public schools. Using data from 2018-2020, the research investigates whether changes in school funding correlate with changes in academic performance.