Architectural Digest
Open Door:
Inside Celebrity Homes
Comment Section Analysis
Introduction
In my project, I will take a closer look at Architectural Digest‘s popular series “Open Door: Inside Celebrity Homes“, where celebrities give tours of their homes. As an architecture admirer, my goal is to explore the standout features of these homes and understand how they resonate with viewers.
I aim to identify recurring themes and sentiments within these videos through extensive analysis. Additionally, I plan to use generative AI to recreate parts of these homes and maybe even evaluate how accurately the essence of the original spaces can be found in the result. I will also be conducting research about generative art, looking into what risks it poses, what advantages it offers and the importance of using it mindfully and transparently
Research
Architectural Digest
- Identified themes of authenticity vs. staging, examining cases where homes were revealed to include props, rented spaces, or promotional elements.
- Looked at the timing of home tours relative to property sales, suggesting potential promotional motives behind certain episodes.
AI Image Generation
- Understanding AI in Design: Explored how AI tools like Stable Diffusion can reinterpret architectural themes and assist in generating creative, data-driven visuals.
- Ethical Considerations: Analyzed the potential risks of misrepresentation in AI-generated visuals, emphasising the importance of transparency and ethical use in creative projects.
- Human-AI Collaboration: Researched the balance between AI’s efficiency in automating tasks and the irreplaceable role of human creativity and critical thinking.
Development
Data Collection & Cleaning Process
- Gathered structured video data and thousands of YouTube comments using the YouTube Data API, focusing on Architectural Digest’s “Open Door” series.
- Overcame challenges such as pagination limits, inconsistent formats, and multilingual comments by developing a systematic cleaning process. Techniques included tokenisation, emoji handling, removing duplicate/ too short entries and punctuation.
- Repeated the same process with each video description and standardised the format.
Topic Modelling
- Before diving into the analysis, I generated word clouds to quickly identify recurring themes in the comments, getting an initial sense of frequently used words.
- Performed Latent Dirichlet Allocation (LDA) for topic modelling to uncover thematic trends in the comments, such as main architectural features or recurring design preferences.
- Integrated video descriptions into the analysis to identify discrepancies or alignments between viewer perceptions and how the videos are marketed.
- Conducted additional cleaning and formatting steps along the way.
Sentiment Analysis
- Conducted sentiment analysis using the NRC Emotion Lexicon to extract emotional trends from comments, categorising responses into emotions such as joy, surprise and sadness.
Reflections
This project taught me the importance of adaptability and iteration. From handling complex datasets to refining AI outputs, I strengthened my technical and creative problem-solving skills. Additionally, it highlighted the balance between leveraging AI tools and maintaining ethical, human-centered design practices. I learned the importance of flexibility in project planning, especially when working with unpredictable datasets.
Explore more of my Data-driven Work
