Word Cloud Plume: phrases, sentiment, comparison and shapes

These recipes reuse the Topics and Reviews tables of the getting started guide. Recipes marked ★ use a premium feature: it works without a key, with the "Premium preview" watermark.

How do I exclude words that come from a table?

You maintain a list of words to ignore (brand names, "hello", "thanks") in an Excluded words table.

1. Create the measure:

Excluded word list =
CONCATENATEX ( 'Excluded words', 'Excluded words'[Word], ";" )   -- "hello;thanks;plume"

2. Drop it into the visual's Excluded words field.
3. Result: these words disappear, in addition to those of Format › Words › Other words to exclude and the built-in stop words. Case and accents are ignored.

How do I count answers rather than words ★?

In a survey, a respondent who writes "delay" three times should count once.

1. Format › Words › Count: Once per row (answer, comment) ★.
2. Result: the size of a word = number of answers that use it; the tooltip says "Rows using it".

How do I merge "order" and "orders" ★?

1. Format › Words › Merge singular and plural ★: on.
2. Result: simple plurals (‑s, ‑x, ‑ies → ‑y) join the singular when it exists; the most frequent form is shown.

How do I see two-word phrases ★?

"customer service" says more than "customer" and "service" apart.

1. Put a text column in Words or text.
2. Format › Words › Phrases: Two-word phrases (or Words and two-word phrases, or Three-word phrases for "delivery lead time").
3. Result: a stop word never starts or ends a phrase, and a phrase never crosses punctuation ("fast, service" is not a phrase).

How do I color words by review sentiment ★?

Your Reviews table has a Sentiment column from −1 (negative) to 1 (positive), for example computed with AI Insights or Power Query.

1. Create the measure:

Average sentiment =
AVERAGE ( Reviews[Sentiment] )   -- average of the reviews in the context

2. Drop it into Color. The visual switches to Gradient by the “Color” field.
3. Format › Data colors › Gradient: Diverging (around the middle).
4. Result: each word takes the average of the reviews that use it; words of positive reviews on one side of the gradient, negative ones on the other, the legend gives the limits.

How do I show rising and falling words ★?

1. Create the measure for last year:

Mentions last year =
CALCULATE ( [Number of mentions], SAMEPERIODLASTYEAR ( 'Date'[Date] ) )   -- same period, one year earlier

2. Drop it into Comparison.
3. Result: ▲ in green for rising words, ▼ in red for falling ones, gray for changes under 5% (Format › Comparison ★ › Negligible change (± %)), ✦ for new words. The tooltip gives the previous value and the change in %. A click on "Rising ▲" in the legend selects every rising word.
4. For complaints, turn on A decrease is favorable: the colors swap.

How do I give the cloud the shape of my logo ★?

1. Open your SVG logo in a text editor and copy the whole svg text.
2. Format › Shape › Shape: Custom SVG, then paste it into Shape SVG.
3. Result: the words fill the solid parts of the logo. The SVG is cleaned (scripts, images and web links removed); only its drawing is used.
4. Variant: a measure that returns a different SVG depending on the filter (the logo of the brand picked in a slicer), in the Shape (SVG) field.

How do I fill the whole shape ★?

With few words, the shape is hard to recognize.

1. Format › Layout › Fill ★: turn on Fill the shape with repeated words.
2. Result: once every word is placed, smaller copies fill the remaining room. Repeated words opacity pushes them to the background; they are neither clickable nor read by screen readers.

How do I make the cloud change month by month ★?

1. Drop a month (or year) column into Play axis ★.
2. A play bar appears under the cloud: ▶, slider, speed ×0.5 to ×4. The current value is written in large type.
3. Result: one cloud per month; common words roughly keep their place and glide to the new one, the others appear or fade out. Format › Playback ★ sets the duration of a frame, the loop and the frame shown on opening.

How do I color words from light to dark by size ★?

1. Format › Data colors › Coloring: Gradient by size ★.
2. Format › Word text › Size by weight: Logarithmic (compressed) ★ when a few words crush the others, or By rank ★ for even sizes.
3. Format › Shape › Silhouette in the background ★: on, Silhouette opacity = 25.

Result: big words take the strong end of the gradient, a gradient legend appears under the cloud, and the shape shows in the background. Without a license, Single color, Palette and the Square root scale stay free.

How do I keep word sizes when a slicer filters the cloud?

By default, the largest visible word takes the maximum size: after a filter, a small topic can become huge. To compare sizes from one filter to another, give the visual the largest weight computed without filters.

1. Create this measure, named exactly wordFont.weightMax:

wordFont.weightMax =
CALCULATE (
    MAXX ( VALUES ( Topics[Topic] ), [Number of mentions] ),   -- the largest weight of a topic…
    ALL ( Topics )                                             -- …without any filter on the Topics table
)

2. Drop it in the Dynamic format field: it fills Format › Word text › Weight at the maximum size (empty = automatic).
3. Format › Layout › Size fitting = Fixed sizes.

Result: a topic keeps the same size whatever the filter; a filtered cloud looks smaller instead of being enlarged.

How do I keep hashtags, mentions and emojis?

In reviews or messages, "#sale", "@store" or an emoji count as much as a word.

1. Drop the text column (Reviews[Comment]) in Words or text.
2. Turn on Format › Words › Keep special characters (#, @, emojis).

Result: hashtags, mentions, amounts ($20, 20€) and emojis stay words of the cloud instead of losing their sign. Emojis show in color with the Windows Segoe UI Emoji font.

How do I always get the same cloud, or try another one?

The layout is deterministic: same data, same cloud, every time you open the report and in the exported PDF. Format › Layout › Layout variant (1 to 999) offers another arrangement, just as stable.