What Shorewood Said

Where the data came from

A resident built this page with an AI model (Claude) doing the extraction, the tagging, the checks, and the page itself, under a person’s direction and review. Everything on it was extracted from the Village’s published PDF: the 1,459 comments from the table on pages 34 to 102 of the report (pages 36 to 104 of the PDF file), the chart values from the report’s own labels, and the response-count tables that sit under each chart as images. Counts were checked two ways: each table’s answers must add up to its total, and the average computed from the counts must match the average printed on the chart to within 0.1. Where the report’s own numbers did not line up, the notes below say so.

How the comments were tagged

Each comment was read by an AI model (Claude) and tagged with up to several topics from a fixed list of 41, a stance on 13 contested questions where the comment clearly takes a side, and an overall tone. A person directed the work, set the topic list and the tagging rules, and reviewed the output. The topic tags were then cross-checked against keyword searches (for example, every comment containing “meter” was compared with the comments tagged “metered parking”) and the differences reviewed. Some subjects are split across themes on purpose: a comment about taxes may sit under “Property tax levels,” “Willing to pay,” “Fees on the utility bill,” or “Spending & fiscal management” depending on its point, so searching the word “tax” finds more comments than any one theme holds.

What that means for reading the counts. Topics with concrete words behind them, such as parking meters, the DPW facility, the Oakland lot, or leaf pickup, are the most dependable. Tone is a judgment about the comment as a whole, and a comment that praises one thing and criticizes another may land in “mixed” or in either camp. Stance is recorded only where a comment plainly takes a side, so stance counts understate how many people care about a question. No formal accuracy audit has been done, so read a count of 164 as “about 164,” and read the comments behind any number before quoting it.

The map, and the Census comparison

Street and park geometry comes from OpenStreetMap. The eight survey sections were redrawn by hand from the report’s map, so their edges are approximate. Place mentions are found by matching names in the comment text (for example “Oakland” or “Atwater”), and the matching rule is shown when you tap a place. The Census comparison uses the American Community Survey 5-year estimates for the Village of Shorewood, with the table numbers listed under the chart; age is compared among adults only, since only adults answered the survey.

Names

The Village published all comments verbatim, including a few that name individual residents, staff, and officials. Here, personal names are replaced with a role in brackets, such as “[a trustee]”. Business names, streets, and public figures in the media are unchanged.

Where the report’s own numbers don’t line up

While extracting the data, a few places turned up where the report’s charts, tables, and labels disagree with each other. They may be layout or labeling slips in the report rather than problems with the survey itself. They are listed here because they affect a few numbers on this page and explain some choices made in building it.