Results and Observations
National patent offices (we examined EPO, WIPO, JPO, SIPO, KIPO, USPTO, and IP Australia) generally provide only their own patent data, but with quite varied interfaces.
By the other public sites and the commercial sites, the data coverage is generally within a minimal set: US, EP, PCT, sometimes JP and INPADOC. Other occasionally included countries are typically European countries – notably the United Kingdom, Germany, and France. Questel-Orbit appeared to have the largest set of countries covered, although coverage from some of these countries was admittedly inconsistent. The commonality of datasets is not surprising given that there are a limited number of organizations that produce the data; the providers here most likely procure their data from these few producers.
Factors that set apart these surveyed providers comprise input format, the results format and the ease of using the site. Although less important, convenience factors such as the speed of results being returned and navigation played a role.
As discussed below, most of the providers have an input interface that is at least usable by an examiner. Most of the sites use “fill-in the blanks” forms, in which the form contains a set of fields (e.g. assignee, title, abstract, inventor). As an alternative, there may be a blank box that accepts Boolean queries for which the user specifies the field(s) associated with each search term. Refining queries depended on the style of input. Depending on the provider, the user has to return to the form or can directly modify the query that is written out in standard Boolean format. As noted below, some sites retained a search history, making the refinement of previous searches more convenient. With regard to inputs, there was no apparent difference which technology area was being queried, with the following exceptions: design patents often required a different interface, and a user attempting to search chemical structures or biological sequences would be extremely limited in the ability to find prior art read on by the claims.
The biggest difference between providers, and usually the biggest deficiency area, is the output i.e. the results list. Capabilities to reformat and re-sort the list and to choose fields for display can greatly enhance finding the closest prior art where there are many search results. Unfortunately, the lack of transparency and inflexibility of Google is being mirrored by many intellectual property data providers, flexibility being sacrificed for seeming simplicity. The best providers in this area are Delphion and WIPS, and CAMBIA’s Patent Lens.
A few of the commercial data providers also have available non-patent databases, some of which are also available to the public. Among the fee-based and subscription-based sites, DialogPro and STNeasy have a limited number of databases (about 100) and Dialog and Lexis have many hundreds each.
Whether these databases are sufficient for the examiner searching non-patent literature depends upon the technology area. For applications with claims to specific biological sequences, the searcher needs GenBank at minimum, which can be searched by the public cost-free through NCBI (to which CAMBIA provides a link) and optimally Derwent’s gene sequence bank that also has sequences published in patents. For biotechnology (other than sequences), PubMed4 and BIOSIS are important non-patent databases. Several of the commercial providers charge fees for searching these databases, and the searches are usually separate from the patent data.
For chemical patent applications, a database where chemical structures are searchable is often necessary; STN, in connection with CAS, does have such a database (but not in STNeasy). CAMBIA will investigate the possibility of obtaining this data for public good use.
The mechanical arts field typically relies mainly on patents for prior art. For computer and software technology, there is ongoing debate about where and how to do prior art searches of non-patent literature.
Most of the comments related to utility patent document searching also apply to design patents. Only one of the databases (PatentCafe) allowed the user to choose only design patents for the search. As design classifications are the primary search terms, its unclear that this sub-division is much help because utility patents that may have a design element will be excluded from the results list, and vice versa. This feature may even be detrimental, as design patents can be valid prior art against utility patents and utility patents can be cited against design patents. Searching for design patent prior art in the non-patent literature is also gravely limited by reliance on classification for most design application examination.
4CAMBIA provides a link to PubMed and will investigate the development of APIs to this and other NCBI databases to make the searches more seamless.