paperchart — recall / parity chart
A thin dot plot with one row per query. Each row has three overlapping dots at the same x position; two are neutral, one is accented, and the caption to the right confirms that all three engines returned the same set.
When to use
- Search engine / retrieval model parity claims.
- "all three variants return the same top-k" stories.
- Any time you need to show a Jaccard similarity of 1.00 without resorting to a table of numbers.
Visual rules
- One row per test case. The label sits to the left, right-aligned.
- Three circles at the same x coordinate, vertically offset by 6 px each, so they read as stacked. Two neutral, one accented.
- A thin
#E6DCCEtrack line connects the left axis to the dot so the eye lands on the right row. - The right-side caption (e.g.
5 hits · Jaccard 1.00 across all three engines) confirms equality in plain English. - Legend at the bottom lists the three variants with a single colour dot each; no lines, no boxes.
How to use
- Edit
src/charts/RecallChart.tsx. - Replace the
QUERIESarray with your test cases. - Update the legend labels if your variants are not labelled "engine A / B / C".
npm run dev, open/#/recall.npm run snapto render toout/recall.png.