Narrative Sentiment Tracker
A research tool for tracking emotional valence in public narratives over time, enabling researchers to visualize how sentiment shifts across narrative arcs in political and social discourse.
Research Question
How does sentiment shift across narrative arcs in political discourse?
Data Source
Public social media datasets (anonymized)
Limitations
English-language corpus only, sentiment models may miss cultural nuance
Overview
The Narrative Sentiment Tracker (NST) is an experimental research tool designed to measure and visualize how emotional valence evolves across the arc of public narratives. Unlike static sentiment analysis, which scores individual documents in isolation, the NST maps sentiment trajectories — the shape of emotional change as a story, debate, or political event unfolds over time.
The tool addresses a fundamental gap in discourse analysis. Most sentiment tools treat texts as independent data points. But public narratives are sequential and cumulative: the emotional weight of a political speech depends on what came before it, and the impact of a news cycle is shaped by the arc it follows. The NST captures these dynamics by modeling sentiment as a time series and identifying inflection points where emotional tone shifts significantly.
How It Works
Users provide a corpus of time-stamped texts — tweets, news articles, parliamentary statements, or any chronologically ordered discourse. The NST processes this input through three stages.
First, each text unit is scored for emotional valence using VADER sentiment analysis, supplemented by spaCy's linguistic features to account for negation, intensifiers, and rhetorical structure. Second, scores are aggregated into temporal windows (configurable by the user) and smoothed to reveal underlying trends rather than noise. Third, the tool applies change-point detection algorithms to identify moments of significant sentiment shift — the points where a narrative turns.
The Streamlit interface provides interactive visualizations that allow researchers to explore sentiment trajectories, zoom into specific periods, and overlay external events for context. Annotated inflection points are highlighted with confidence intervals.
Research Applications
The NST has been used in pilot studies examining sentiment arcs in electoral campaigns, public health communications during crisis events, and legislative debates on contentious social issues. Early findings suggest that narrative sentiment trajectories follow a limited set of recurring shapes, echoing literary narrative theory. Identifying these shapes early in a discourse cycle may help practitioners anticipate emotional escalation and design timely interventions.
The tool is currently in beta. We welcome feedback from researchers working in computational social science, political communication, and media studies.