Computational Narratives of Brazilian Favelas
Key Words
Natural Language Processing, Machine Learning, Urban Planning, Brazilian Favelas
Date
Dec 2024
This project explores the potential of computational design to analyze and visualize complex urban and social dynamics through Natural Language Processing (NLP). Using Janice Perlman’s seminal work, Favela: Four Decades of Living on the Edge in Rio de Janeiro, as a primary data source, the project applies advanced text analysis techniques like sentiment analysis and topic modeling to uncover hidden patterns and themes in the evolution of Brazilian favelas. These insights are then translated into compelling visualizations to challenge stereotypes and foster understanding of favela communities.
Research Background
My deep interest in computational design stems from a desire to combine quantitative and creative insights to drive meaningful change. Through a combination of social justice courses and my Media Arts and Design coursework, I’ve been drawn to exploring social and urban issues, with Brazilian favelas as a particularly compelling focus. These courses introduced me to the power of creative computing techniques such as machine learning, NLP, and VR as tools for storytelling and change-making. By leveraging these mediums, I aim to bridge the gap between data-driven analysis and emotional engagement, fostering a deeper understanding of marginalized communities and their resilience.
This project builds upon extensive research in both the social and technical domains. The foundational text is Janice Perlman’s Favela: Four Decades of Living on the Edge in Rio de Janeiro, which I am analyzing to uncover evolving themes within Brazilian favelas. Supplementing this, I have explored related news articles to contextualize historical patterns within contemporary narratives. From a technical perspective, I’ve delved into published computational design papers, gaining insights into the application of machine learning and NLP techniques for urban data analysis. Additionally, I have researched VR technology’s role in urban landscape improvement, particularly its potential for immersive storytelling and audience engagement. These elements together form the basis for the innovative integration of computational analysis and creative visualization in this project.
Pipeline
----> analyze the book using NLP methods
----> NLTK & sentiment analysis & Bert topic modeling
----> gain NLP insights
----> data visualizations
----> download google street view images
-----> 3D modeling in Unity
-----> build favela scenes in the VR environment
What is the book about?
Favela: Four Decades of Living on the Edge in Rio de Janeiro by Janice Perlman is a comprehensive longitudinal study that examines the lives of residents in Rio's favelas over a span of 40 years. Building upon her earlier work, The Myth of Marginality (1976), Perlman revisits the communities she first studied in the late 1960s, conducting nearly 2,500 interviews with original participants and their descendants to assess changes in poverty, social mobility, and urban policy impacts.
NLP Results
NLTK analysis
"Mobility and 'upward' are highly correlated → suggesting the aspiration of favela residents to move up in life."
theme: Drug
theme: Violence
NLTK sentiment analysis (news)
Has the environment of Rio’s favelas improved after the Rio Olympics? (2016 Rio Olympics)
.png)
.png)
.png)
.png)
.png)
.png)
Bert Topic Modeling
This BERTopic analysis reveals a diverse and layered discourse surrounding life in Rio’s favelas. The model identifies 72 distinct topics, which are organized into broader thematic categories such as city dynamics, population groups, infrastructure, daily life, and social issues. Topics like “mobility” and “upward pattern” highlight aspirations for social advancement, while others such as “police & military,” “violence,” and “favela removals” underscore structural challenges and state interventions. Social identity topics including women, youth, and Black communities indicate demographic-specific concerns, whereas infrastructure-related themes such as electricity and water point to basic service access. The presence of terms like “citizenship,” “inequality,” and “marginality” within the social issues cluster reflects systemic struggles tied to human rights and socio-economic divides. Together, the model captures both the material and symbolic dimensions of favela life, illustrating how residents navigate conditions of precarity while striving for dignity, inclusion, and progress.
.png)
Sankey Diagram
Sentence Anchors
Google Street View and 3D Modeling
I selected a single image from Favela: Four Decades of Living on the Edge in Rio de Janeiro and located its real-world counterpart using Google Street View. This site was chosen for its representativeness of the broader favela landscape. I transformed the image into a field of colored dots and built a 3D point cloud model from it. This technique, inspired by my NLP topic modeling results that emphasized “mobility” as a dominant theme, serves to visualize movement and depth—capturing both the physical elevation of the favela and the layered social mobility that defines life within it.
3D Modeling Processing
.png)
.png)
.png)
.png)
.png)