Summary

  • USPTO's fiscal 2000 report dates its older patent image expansion to December 1999: 4,204,863 patents from 1790–1975 became available as page images, searchable by number and current classification.
  • Patents from 1976 onward could be searched through full-text fields. The archive's reach and a keyword query's reach had different starting dates.

A very large arrival, with a small condition

More than four million old patents came onto the web with an unusually useful piece of baggage: their pages. Readers could inspect the documents themselves. To find an older one, the service offered entrances through its number and its place in the classification system. The archive had moved closer; knowing one's way around the catalogue remained an advantage.

United States Patent and Trademark Office recorded the expansion in its fiscal 2000 report. In December 1999, the database grew to more than 6.5 million patents reaching back to 1790. The older addition comprised full-page images for 4,204,863 patents issued through 1975.

Then came the distinction that gives this impressive total its shape. Those older records were searchable by patent number and current patent classification. For patents issued from 1976 onward, the service offered full-text fields, including classification. The search box had a historical frontier.

The words arrived in stages

The public service had already been widening for several years. USPTO's technical report to WIPO dates its web bibliography and abstract database to November 1995. That collection covered patents back to January 1976 and was updated weekly, usually on Tuesday, the issue day. The raw data were also available by FTP on the same day.

In November 1998, the agency added searchable full text for that period. Its 1999 annual report then records the formal unveiling of new Internet databases in April 1999, including text and images for more than two million patents from 1976 onward. Bibliographic entries, searchable wording and page images were separate pieces of the public offering, assembled over time.

The December expansion carried the images much further into the past. The keyword search boundary stayed at 1976. A reader following a known patent number could reach an old page that a search through newer full text would never have examined.

A photograph cannot do the index's job

A full-page image preserves the document's appearance, including drawings and the arrangement of text. A search service also needs a way to identify which page to fetch. The distinction was familiar in USPTO's other products. The technical report described its USAPat CD-ROM as a document-delivery system: facsimile pages, including drawings, retrieved by document number through a cumulative index.

That is a good tool when the number is known. Discovery asks more of the catalogue. Someone approaching an older subject through its classification had to identify the relevant category of invention; someone with a document number could go directly to the record. The words a reader might type were only one possible route through this material, and their useful range depended on the collection searched.

The report also described work on optical character recognition for older patents in the section on the examiners' search file. Loading converted text into the internal search engine was a further planned step. That work had its own audience and timetable. The public service's documented search fields tell us what the web visitor could actually ask it to do.

The catalogue still had people nearby

Physical research support was growing alongside the web service. The 1999 annual report records two new Patent and Trademark Depository Libraries, designated that May, bringing the network to 85. Millions of online pages and a network of places to work with patent information existed together.

The interesting history sits in that combination. Delivery over the Internet made an immense collection easier to reach. Numbers, classifications, text preparation and finding aids determined how a reader could discover something inside it. The page count announced a great deal of progress; the index explained how to use it.

A result list was therefore a record of a particular search through a particular set of fields. Before celebrating a clean slate, it paid to notice which years the words had been allowed to visit. Four million documents can be present and perfectly readable while waiting for a rather different introduction.