G — Physics

Reading this classification

Patent classification exists because keyword search fails in a specific way on patents: the same invention is described in different words by different drafters, sometimes deliberately. A class is assigned by an examiner on the basis of what the invention is, so it groups filings that share no vocabulary at all. Finding the right class is usually a faster route into a field than any search term.

The Cooperative Patent Classification is the scheme in current use, developed jointly by the USPTO and the European Patent Office. It divides all of technology into nine sections, each subdivided down to fine groups. A code such as B64C29/0008 reads as section B, class 64, subclass C, then the specific group. This archive browses at section level, because the nine section titles are official and unambiguous while subclass titles are not something we hold — inventing plausible ones would corrupt exactly what a reference archive exists to provide. Full codes are printed on each record.

One document commonly carries several codes. Classification is not a partition: an optical measurement technique used on a chemical process legitimately belongs in more than one place, so finding a patent under one section does not mean the others are irrelevant.

Classification is applied by the office rather than by the applicant, which is what makes it more reliable than a title for finding related work. It can also shift: schemes are revised and documents reclassified in bulk, so a class cited in older literature may no longer exist under that name.

Classified browsing is powerful and incomplete. An application whose disclosure spans several fields may be classified narrowly by its claims and missed by a search of the others. Serious prior-art work combines class browsing with keyword and citation searching, each catching what the others miss.