Urban Sentiment Study on Downtown Pittsburgh
NLP | Python | Data Visualization
Independent Study
Instructor:
Kristen Kurland
This project explores how everyday public perception can inform urban planning decisions in Downtown Pittsburgh. Using social media posts as a form of crowdsourced feedback, I applied natural language processing (NLP) and Python-based text analysis to examine how residents describe their neighborhoods, streets, and daily experiences. By transforming unstructured text into structured sentiment data, the study treats online discourse as an informal but continuous layer of community input, complementing traditional surveys, public meetings, and engagement processes.
X
X captures immediate, event-driven reactions to Downtown Pittsburgh. The charts show an overall decline in posting volume in recent years, but sharper short-term spikes in both positive and negative sentiment tied to specific moments. Positive keywords cluster around events, markets, and public spaces, while negative language centers on crime, safety, and street conditions. Together, the data suggests X functions as a real-time pulse of the city, useful for monitoring public response to incidents, programming, and short-term planning actions.
X(Twitter) Positive Sentiment Posts Word Cloud
X(Twitter) Negative Sentiment Posts Word Cloud
Reddit
Reddit reflects slower, more sustained community discussions rather than instant reactions. In contrast to X, the trend lines show growing activity and higher overall engagement over time, with longer-term sentiment patterns. Positive posts emphasize neighborhood pride and everyday livability, while negative conversations focus on structural concerns such as safety, housing, and local services. This platform acts more like an ongoing public forum, helping planners understand deeper community narratives and persistent quality-of-life issues.
Reddit Positive Sentiment Posts Word Cloud
Reddit Negative Sentiment Posts Word Cloud
X (Twitter): Real-Time Sentiment Signals
Z-score trends show that sentiment on X fluctuates sharply month to month, with distinct positive and negative spikes tied to specific events. In recent years, both posting volume and positive sentiment gradually decline, while negative peaks appear around isolated incidents. Compared to other sources, X behaves as a short-term indicator rather than a stable trend, making it useful for tracking immediate public reactions and evaluating the impact of near-term disruptions or programming.
Reddit: Long-Term Community Discourse
Reddit displays a steadier but upward trajectory in both activity and sentiment intensity. The Z-score curves show fewer abrupt spikes but more sustained periods of positive or negative discussion, particularly after 2023. This suggests deeper, ongoing conversations about neighborhood conditions and quality-of-life issues. As a result, Reddit functions less as a momentary signal and more as a barometer of persistent community concerns.
Sentiment Trends (Z-Score Comparison) Summary
Standardizing sentiment scores using Z-scores allows patterns across platforms to be compared on the same scale, highlighting when discussion intensity deviates from the norm. While X captures short-term shocks, Reddit and local media reveal longer-term shifts in perception. Together, these signals provide planners with both immediate feedback and structural insight, supporting more responsive and evidence-based urban decision-making.
Through sentiment classification, temporal comparison, and word-frequency visualization, the analysis highlights recurring urban themes tied to quality of life. Positive sentiment clusters around parks, events, walkability, and local businesses, while negative language frequently references safety concerns, traffic, and street conditions. These patterns reveal how people emotionally respond to specific places and issues, helping identify where planning interventions may be most needed and which neighborhood assets are most valued.
By bridging computational methods with planning research, the project demonstrates how large-scale text data can support evidence-based decision-making and more responsive urban policy. This approach enables planners to monitor public sentiment over time, evaluate the social impact of investments, and surface voices that may not participate in formal engagement channels. Ultimately, the work positions data analytics as a practical tool for community-centered planning and more equitable urban development.