An Aura of Words

The course's idea that stuck was small data: against the reflex to reach for big datasets, it argues for situated, citizen-generated data, qualitative and grounded in lived experience, and meaningful precisely because it is not produced to satisfy a report. A Google Review is exactly that kind of small data.

So the project's whole stance is a debate, not a verdict. It does not ask whether the parks are good or bad. It asks whose experience counts in how we understand public space, and whether the collective voice of citizens, once made legible, can say something the official channels cannot.

The raw material is Google Reviews of five Lugano parks, gathered with a custom Python scraper that loads each park's reviews and pulls the text, the star rating, the date, and the reviewer name. Only reviews with written text were kept, in whatever language they were left in, mostly Italian, with German, English, and French in the mix.

The five parks are deliberately unequal, because that is how attention really falls: Parco Ciani alone accounts for more than 1'600 reviewed words, while little Parco Lambertenghi contributes 65. We kept that imbalance rather than normalizing it away, because the volume of what people choose to say about a place is itself part of the portrait.

Every meaningful word is sorted into one of six lenses: the visitor's inner state, the sensory environment, the actions people take, the social context, the physical infrastructure, and tension or complaint. Functional words are ignored; only the words that carry real descriptive weight are kept.

The encoding is a hand-curated lexicon matched word by word across each park's reviews. Each occurrence also records the other meaningful words that appear alongside it in the same review, so the data captures not just which words are used, but which ones tend to appear together.

Each park's aura is an organic, color-blended blob where the area of each color is the proportional weight of that category in its reviews, with a subtle breathing motion so it reads as a living portrait rather than a fixed chart. Color is the primary encoding, so a reader recognizes a category instantly, without labels.

What gives each aura its character is how the lenses mix. People rarely describe a park in one register; a single review often moves between feeling, sensory detail, and complaint in the same breath. The aura is built to show that blend, not just which lens is largest, but which ones tend to occur together, which is what makes one park's portrait recognizably different from another's.

The interface is a guided scrollytelling journey. It opens over a drift of real review snippets, lays out the six-category framework, then runs a methodology lab where you watch a single review get taken apart word by word, each meaningful term lighting up in its category color as a running tally builds the aura in front of you.

From there it opens out. A map places the five auras on Lugano, sized by review volume; a comparison view sets their chromatic signatures side by side; and a per-park word map shows every categorized word as a color-coded network, so you can see how themes interconnect. It ends by handing the method to the reader: a prompt asks what your ideal park would be, and as you type, your words are classified live against the same six lenses. You stop being an audience and become another voice in the data.