Word Counter

Counts words, characters (with and without spaces), sentences, and paragraphs in a text input. Word boundary detection uses Unicode-aware whitespace and punctuation rules, handling languages with non-space delimiters (Chinese, Japanese) by counting characters as proxy.

Reading-time estimate at 250 WPM and speaking-time estimate at 150 WPM are reported alongside. Sentence detection uses end-punctuation heuristics that occasionally misfire on abbreviations (Mr., Dr., etc.).

Word definition is the source of most counting discrepancies. The standard rule treats a word as a maximal sequence of non-whitespace characters separated by whitespace or punctuation. Hyphenated compounds like "mother-in-law" count as one word; contractions like "don't" count as one word; em-dash separated phrases like "the cat—the black one—jumped" count the bracketed words separately. Microsoft Word, Google Docs, and most text editors implement this rule with minor variations; the counter matches their behavior closely.

Sentence detection runs into the abbreviation problem. The naive rule "split on . ! ?" produces wrong counts on text like "Dr. Smith arrived. The meeting started." which is one period inside an abbreviation followed by one terminating period — two visible periods but one sentence-ending. Better detection requires either a list of abbreviations to ignore (Mr., Mrs., Dr., Prof., U.S., etc.) or a small statistical model that considers context. The counter uses an abbreviation list approach, which catches common cases but misses rare ones.

A worked example for typical English prose. Input: "The quick brown fox jumps over the lazy dog. It was a calm afternoon. The fox decided to nap." Word count: 17. Character count (with spaces): 90. Character count (without spaces): 73. Sentence count: 3. Paragraph count: 1. Reading time at 250 WPM: 4 seconds (rounds up to "less than a minute"). Speaking time at 150 WPM: 7 seconds. The metrics give a sense of the text's bulk and ease of consumption.

CJK languages (Chinese, Japanese, Korean) require different handling. These languages do not use spaces between words; word boundaries require linguistic analysis. The character count is the conventional metric in those languages anyway — Chinese textbooks specify lengths in characters (字), Japanese specify in characters (字) for ideographs and morae for kana. The counter falls back to character count for these scripts. Mixed-language documents (English with embedded Japanese) are counted segment-by-segment.

Limitations: HTML tags are counted as words by default because they contain non-space characters. Pasting "<strong>important</strong>" in produces a count of 3 words (<strong>, important, </strong>), which is rarely what the user wants. The strip-HTML mode removes tags before counting; the counter offers this as a toggle. Markdown content has similar issues — ** counts as part of a word boundary or not, depending on the parser. For accurate prose word counts of formatted content, render to plain text first and count the result.