Whitespace splitting misses languages without spaces and mishandles punctuation around words. WordCount uses Intl.Segmenter with the selected analysis locale and counts only word-like segments. The chosen locale is part of the result contract.
Choose the language of the text
The page language controls labels; the analysis language controls word and sentence boundaries. Select the language that best matches the text instead of assuming the interface language describes its contents. Mixed-language text remains an approximation because one segmenter locale covers the whole input.
Punctuation is not automatically a word
Segmenter returns spaces and punctuation as segments too, so WordCount requires the platform isWordLike flag. Standalone punctuation does not increase the word total. Numbers can be word-like, which matches many editing tools but should be checked against a specific academic or publishing rule.
Frequency uses conservative normalization
Frequency terms are normalized to NFC and lowercased with the selected locale. Accents are preserved, so resume and résumé remain different. The table is capped at ten items and never enters copied summaries or analytics because the terms are derived directly from private input.