Interactive explainer

The Vocabulary Gap: Why the Same Invention Has a Dozen Names

A “hoverboard,” a “self-balancing scooter,” a “personal mobility device,” a 1990s “human transporter,” 自平衡车, and a CPC code — all the same invention. Type the one word you’d search for and watch how much of the prior art it never reaches.

One invention, many names
90%
of how this art is described would be missed
Your term matches 1 of 10 phrasings.

Each greyed node is a real way this invention is described that your keyword never reaches.

concept(the meaning)hoverboardself-balancing scooterSegway-type vehiclepersonal mobility deviceself-stabilizing vehicleinverted-pendulum transporterPMD自平衡车human transporterB62K 11/007
EverydayTechnical jargonAcronymOther languageOlder termClassification

The same idea wears a dozen disguises

Keyword search rests on a quiet assumption: that the prior art uses the same words you do. It almost never does. A single invention is described with everyday synonyms, technical jargon, acronyms, translations, dated terminology from when it was first filed, and classification codes that contain no natural-language words at all. Every one of those is a place decisive prior art can hide.

Why the missed phrasings are the dangerous ones

The references most likely to invalidate a patent are frequently the ones written years earlier, in another field, or in another language — precisely the phrasings a keyword search drops. An examiner who searched “hoverboard” would sail right past a 1998 patent for a “self-stabilizing human transport vehicle,” even though it describes the same machine. The gap isn’t an edge case; it is the typical case, and it is invisible — you don’t see the results you never retrieved.

Flip to Semantic. Instead of matching letters, semantic search matches meaning — so every phrasing collapses into a single concept and the miss rate falls to zero. That’s the bridge across the vocabulary gap, and it’s why a thorough modern search pairs keyword precision with semantic recall.

What to do about it

You can’t brute-force a vocabulary gap with more synonyms — you will never think of the dated term, the foreign translation, or the framing from an adjacent field. The durable fix is to search on concepts: let the system find everything that means what you described, then narrow with the precise terms you do know. See the companion explainer, How AI Prior-Art Search Actually Works, for the mechanism behind it, and Anatomy of a Patent Claim for why finding that one oddly-worded reference can decide a case.

A note on this demo: the phrasings shown are representative, not exhaustive — real inventions have far more. The point is the shape of the problem, not a complete thesaurus.

Search meaning, not vocabulary

PatentScan finds prior art across synonyms, jargon, and languages — so the reference that defeats a patent doesn’t slip through just because it used different words.