I build computational methods that serve social and organizational decision-making. My work sits at the intersection of NLP, machine learning, and computational social science, with applied threads in communication, safety/health, and public policy. This page complements my Projects page by grouping publications and artifacts under the research lines they advance.

Themes & Questions

Affective meaning in digital communication

How do people signal and interpret emotions in text-based settings, and how can models respect sociocultural context?

Data-driven safety & health

When do sensors, analytics, and feedback systems change behavior and reduce risk?

Computational policy analytics

What signals forecast political momentum, and how stable are relationships between money, polling, and campaign events?

Expectations in socio-economic systems

How do modeling choices about expectations affect stability, identifiability, and interpretation in macro-style systems?

Publications by Research Line

Affective Meaning & NLP for Messaging

I combine transformer representations with affect control theory to better measure context-dependent emotion in chat and short text. The goal is interpretable affective signals that are useful for support agents, conversational systems, and social inquiry.

Project result / artifact to visualize This line introduced BERTNN, a method for estimating affective meanings and expanding affective lexicons from contextual language use. Related work maps Emoji2vec representations into affective space to model emotional transitions in messaging.
Contextual Embeddings in Sociological Research: Expanding the Analysis of Sentiment and Social Dynamics.
Sociological Methodology, 2024.
cite
Adapting Online Messaging Based on Emotional State.
UMAP ’21.
pdf
cite
How emoji and word embedding helps to unveil emotional transitions during online messaging.
IEEE SysCon ’21.
pdf
cite
code

Safety, Sensors & Behavior Change

From RFID+haptics for collision warnings to telematics-guided eco-driving and sleep technology feasibility, I study how analytics and design can reduce risk and support healthier behavior in the wild.

Project result / artifact to visualize The RFID collision-warning study reported collision prediction with less than 14% false alarms. In the Mindful Driving project, linked UVA coverage reported nearly 6% fuel-economy improvement, more than 23 gallons of fuel saved per vehicle, and 457 pounds of annual greenhouse-gas reductions per vehicle.
Collision Prediction and Prevention in Contact Sports Using RFID Tags and Haptic Feedback.
AHFE Wearable & Assistive Technology, 2021.
cite
Safe and Sustainable Fleet Management with Data Analytics and Training.
Systems and Information Engineering Design Symposium (SIEDS), 2021.
coverage
Preliminary feasibility of technology use in an internet-delivered intervention: Improving sleep in older adults with mild cognitive impairment.
Alzheimer’s & Dementia, conference abstract, 2020.
pdf
cite

Computational Policy Analytics

Using time-segmented models such as joinpoint regression, I examine how fundraising and polling co-evolve in U.S. primary campaigns and what those dynamics imply for forecasting, momentum, and resource allocation.

Project result / artifact to visualize This work compares polling and financial contributions during the 2020 Democratic primaries and uses change-point analysis to identify moments when campaign trajectories shift, including shifts associated with debate performance and candidate support.
A Tale of Two Metrics: Polling and Financial Contributions as a Measure of Performance.
IEEE SysCon ’21.
cite
preprint

Expectations & Macro-Style Systems

My early work studied expectation formation and solution properties in macro-style systems, with emphasis on stability, interpretability, and plausible micro-foundations.

Project result / artifact to visualize This line connects systems modeling, rational expectations, and multi-agent simulation. A strong visual here would help readers quickly see the relationship between assumptions about expectations and system-level behavior.
Why the determinacy condition is a weak criterion in rational expectations models.
International Conference on Business and Economics Research, 2010.
A predictive multi-agent approach to model systems with linear rational expectations.
First Iranian Economic Conference, 2011.

Methods & Tooling

  • NLP: contextual embeddings, BERT-family models, lexicon expansion, sequence modeling for affect.
  • ML: regression/classification, clustering, time-series segmentation, subgroup-aware evaluation.
  • Sensing & Systems: telematics analytics, RFID localization, dashboarding, training feedback.
  • Policy analytics: campaign-performance signals, joinpoint regression, comparative trend analysis.
  • Open materials: selected code and preprints linked above; additional items available on request.

Impact & Collaboration

  • Mindful Driving: supported by a Jefferson Trust grant; cross-unit work with operations and fleet partners; linked coverage reports improved fuel economy and greenhouse-gas reductions from the deployed software comparison.
  • Affective NLP: published in a sociological methods venue, reflecting interdisciplinary work across NLP, social psychology, and computational social science.
  • Collaborations across sociology, economics, kinesiology, nursing, political science, data science, business, and engineering.

Pointers

Google Scholar
Projects
About