Interactive viewer · t-SNE / UMAP · Walk-through · Math & statistics · Industry

Semantic Universe Explorer — SUE Semiconductor Equipment

Selection

View

3D projection 2D projection

Projection

PCA t-SNE UMAP

All four settings side-by-side: t-SNE / UMAP grid.

Filter first (category, company, digits…), then re-project: the visible subset is treated as the whole dataset and gets its own best-fit axes.

Color by

Content type (12 categories) Company Prose→Tabular score Digit Density of Text Cross-corpus outlierness (rank) In-doc typicality Unsupervised cluster (GMM) Marker shapes carry a second variable

Content-type mode: color = category, shape = company. Swapped: color = company, shape = prose / ambiguous / tabular register (3-D plots offer too few shapes to carry all 12 categories). Untick the box for plain dots.

Filter — content category

Isolate a single one of the twelve recurring Global-Impact-Report content types (CEO letter, Scope emissions, water, workforce, etc.).

Filter — content type

All chunks Only prose (digits < 5%) Only ambiguous (5–18%) Only tabular (digits > 18%)

Or narrow further with a range:

40%

Filter — companies

Filter — outliers only

Show only the top 15% most atypical

Ranks each chunk by mean distance to the four other firms’ centroids; identifies passages atypical of the competitor set.

Selection display

Gray out unrelated points

projection fidelity: computing…

New: interactive t-SNE & UMAP viewers — the same corpus through neighborhood-preserving lenses.

Selected point

Click any point on the plot, or select a featured chunk below.

Featured chunks

Walk-through →