Top expressions (chat n-grams)
The one-, two-, and three-word phrases someone actually uses most — filtered so the answer isn’t just "the" and "a" in every chat.
A naive word count in any language is dominated by function words — articles, prepositions, pronouns — that everybody uses constantly and that say nothing distinctive about anyone. Filtering those out by detected language, and then favoring longer phrases over single words when they score similarly, is what turns a frequency table into something that actually reads like a person’s voice: "amore mio" or "you there" instead of "the" or "che".
Example
A raw word-frequency count might put "the" and "and" at the top of any English chat; filtering stopwords and favoring multi-word phrases surfaces something like "love you" or "on my way" instead.
Frequently asked questions
Why filter by "detected" language instead of one fixed language?+
Because a chat isn’t guaranteed to be in any particular language, and a filter built for the wrong one would let through exactly the function words it’s meant to remove.
Do "media omitted" placeholders ever show up in this list by mistake?+
They shouldn’t — a placeholder for an attachment is recognized as media, not as a phrase someone typed, and is excluded from expression counting entirely.
Related terms
- Signature wordThe single word where the gap between two people’s usage is widest — measured as a rate, not a raw count, so whoever writes more doesn’t win by default.
- Emoji counting (grapheme-aware)Counting emoji correctly means counting whole, rendered symbols — not the raw characters a multi-part emoji is secretly built from.