Independent Project · Discourse Analysis
Source: Grupo Bandeirantes de ComunicaçãoTwo Hours of Debate
Parses electoral debate transcripts into three quantitative metrics per candidate: direct answer rate, proposal density, and confrontation index.
The Problem
Seven candidates for the Senate debated for nearly two hours across 46 speech turns. Viewers were left with common uncertainty: which candidates truly answered questions versus delivered rehearsed rhetoric?
What I Built
I processed the entire debate transcript through my NLP discourse analysis pipeline, extracting three objective scores per candidate: direct answering (did they address the question?), proposal density (did they provide concrete plans or just diagnose?), and confrontation degree (constructive dialogue vs. attacks).
Data Sources
Full verbatim transcript of the São Paulo Senate Debate, Grupo Bandeirantes de Comunicação (Sept 9, 2026).
Outcome
Across all candidates, direct answer ratings averaged 6.9/10 while proposal density hovered at 5.1/10. Voters can explore candidates side-by-side in 15 minutes without re-watching 120 minutes of broadcast footage.
Looking for a tailored solution for your organization?
Client projects remain confidential. These open-source cases demonstrate the exact methodology.