Terminology extractor
Pulls the single-word terms out of a document and hands you a glossary skeleton — sorted by frequency or alphabetically, with inflected forms like “tool” and “tools” grouped into one entry.
Your text never leaves your browser.
Paste a document above to pull out its single-word terms.
| Include in export | Term | Type | Uses | Sentences | Forms found |
|---|
Your text, annotated
Needs attention Worth a look — select a highlight for details.
Findings
How the extractor picks terms
A glossary is not a frequency table. If you want to know which words you lean on, including two- and three-word phrases, use theword frequency & repetition finder instead — it reports every token in the document. This tool answers a narrower question: which single words in this document need an entry someone has to write? So function words, bare numbers, and single letters are dropped before you ever see the list, and what remains is labelled and countable.
Uses and sentences are different signals
A word used eight times inside one paragraph is usually an artifact of that paragraph. The same word used eight times across eight sentences is a running concept, and far likelier to belong in a glossary. Both are shown, and terms tied on raw count are ranked by the spread, so the running concept sorts above the local one.
Acronyms and proper nouns are separated out
All-caps words are marked as acronyms — they usually need a glossary entry more urgently than anything else in the list, and theacronym auditor will tell you which of them your document never expands. Capitalized words are marked as proper nouns, but only when they are capitalized in a position where that was a choice: a word that appears capitalized solely at the start of a sentence is not evidence of anything, so it stays an ordinary term.
Grouping word forms
Group word forms is on by default, because a glossary needs one entry fortool, not two: a document that says tool nine times and tools four times gives you a single entry counted thirteen times, with both forms listed beside it. Untick the box and every spelling becomes its own row again, which is what you want when you are auditing what a document literally contains rather than building a term list from it.
The grouping is dictionary-based, not a stemming algorithm. Stemmers work by chopping suffixes, which is why they turn bus into bu and analysis into analysi. Here every form is looked up in a table of real English word pairs, so analyses groups onto analysis, policies onto policy, children ontochild, and bus is left exactly as it is. Verb forms group too:reviewing and reviewed both land on review.
Acronyms are deliberately exempt. The table is keyed on lowercase words, so sendingIDS through it would file your intrusion-detection acronym under id.
Where the word list comes from
The table is 41,385 form-and-lemma pairs derived fromWordNet 3.0, by way ofspacy-lookups-data, reduced to lowercase alphabetic entries. WordNet is a lexical database built at Princeton University and released under a licence that permits redistribution; the full notice ships with the data at /data/lemmas-en.LICENSE.txt.
It is about 900 KB, fetched in the background while the page loads and cached by your browser afterwards. Nothing waits on it: paste before it arrives and you get an ungrouped list that regroups itself the moment it lands, and if it never lands the tool carries on with each form listed separately. Like everything else here, the lookup runs in your browser — the word list comes down to you, your document never goes up.