Remembering Afghanistan

With Prof. Alexandra Siegel — 2022

I designed and led a bilingual computational study analyzing how people across English- and Arabic-speaking communities discussed the U.S. withdrawal from Afghanistan on social media. Instead of looking only at text content, I modeled behavioral patterns in topic engagement, sentiment polarity, and geographic variation to understand how narratives evolved over time and across linguistic communities.

Most research looks at sentiment or topic modeling in a single language. My work investigates how narratives diverge across languages—how English and Arabic speakers frame the same geopolitical moment differently, and how these differences reflect patterns of polarization, regional identity, and collective memory formation.

Research Question

How do English- and Arabic-language users differ in their topic engagement, sentiment patterns, and geographic reactions surrounding the U.S. withdrawal from Afghanistan—and what do these differences reveal about cross-lingual narrative formation?

Business Impact

Provides actionable insight into cross-market sentiment, information polarization, and regional narrative drift, supporting teams working in trust & safety, global communications strategy, risk analysis, and multilingual content understanding.

What I did

• Built a bilingual NLP pipeline using English and Arabic tokenization, topic modeling, and sentiment scoring tools.
• Collected and analyzed a large cross-lingual dataset of tweets around the withdrawal and one-year anniversary.
• Modeled geographic sentiment variation to identify where and how reactions diverged across regions.
• Compared topic alignment vs. sentiment alignment to uncover mismatches in how different communities framed the event.
• Explored causal patterns between withdrawal milestones and shifts in sentiment polarity.

Why it matters

This study shows how linguistic and geographic context shape public reaction to major geopolitical events. By combining behavioral data with bilingual NLP methods, it provides a scalable framework for analyzing global discourse—critical for organizations monitoring cross-market sentiment, emerging risks, and international information environments.