Choose stop words for a meaningful word cloud

Start with the question you want to answer

A stop-word list removes exact entries before the frequency table is built. It can reduce repeated function words, but it also changes what readers see. The right exclusions for a decorative thank-you card may be unsuitable for a set of product complaints. Keep the purpose of the cloud visible while reviewing the list.

WordShapeLab has an optional original English list, plus a local list you can edit. It is not a universal definition of unimportant language. The frequency counter lets you turn the common list off and add your own exclusions, one complete entry per line. Nothing in that custom field is sent to a server. Test an exclusion list in the counter.

Check no, not and other words that change meaning

The default English list includes no, nor and not. If you count Not clear and Clear together with common stop words removed, clear survives while not disappears. That result counts a literal word; it does not show whether the comments praise clarity. Preserve context before drawing a conclusion from the largest entry.

Clear instructions
Not clear instructions

For a cautious first pass, disable common English stop words and count this invented pair. Then repeat with the option enabled and review the exclusions. If complete meaning matters, use Whole lines as phrases or manually prepare short reviewed phrases. The phrase counter matches entire lines; it does not automatically classify feedback.

Build a small explicit exclusion list

Begin with names, question wording or boilerplate that you have a clear reason to omit. Enter workshop on its own line to remove that exact word in word mode. A two-word custom entry does not remove its two individual words from prose. In phrase mode, the exclusion must match the complete trimmed phrase, including punctuation.

workshop
session
please

With Merge letter case enabled, Workshop and workshop match the same lowercased exclusion. With it disabled, case distinctions are retained. Straight and curly apostrophes remain different characters, and a hyphenated expression is its own token. Check actual spellings in the table instead of assuming similar-looking forms always merge. Read the matching and tokenization rules.

Compare cleaned and original counts

1. Keep an unedited local copy of the source text.
2. Count with common stop words off and both minimum filters set to one. Download that report.
3. Add a short custom exclusion list or enable common English words, then count again.
4. Review the exclusion list and compare changed entries, totals and percentages before using the new CSV.

The occurrence-share denominator changes when stop words are removed. A word can keep exactly the same count while its displayed percentage rises. Minimum count and length filters operate later and use the pre-minimum denominator. Include the filter settings when sharing results so another reader can reproduce the calculation.

Use the cleaned list in the editor

The counter exports only kept entries, with the selected counts as weights. The report preserves minimum-filter omissions and parser exclusions. In the layout editor, Paste text has a separate editable stop-word list. Importing the reviewed weighted CSV avoids applying a new text-cleaning pass, so the chosen counts remain under your control. Import the reviewed CSV.

Stop-word choices do not repair duplicated answers, biased samples, misspellings or synonyms. Avoid silently combining entries that merely seem related. When a small group is involved, remove identifiers before pasting and inspect the report before distributing it: the report can contain excluded source words as well as visible ones. Review local storage and sharing privacy.