Software and business-method patents are a relatively new phenomenon and coincide with the explosion of the Internet in the mid-1990s. U.S. patent class 705 ("Data Processing: Financial, Business Practice, Management, or Cost/Price Determination") currently contains about 13,600 patents (both original and cross-reference classification), of which the vast majority—about 11,200—have issued after 1995; more than 5,500 have issued so far in 2002.

According to Ory Holtzman, product manager for Derwent Web of Software, the USPTO (U.S. Patent and Trademark Office) is currently the most liberal of the major patenting authorities in what it permits in software and business-method patents. Both the USPTO and the EPO (European Patent Office) will grant patents for processes using software. In the U.S., inventors can also obtain patents on the software itself if it is applicable to a practical process, and they can get patents for computer-enabled business methods not yet patentable in Europe (e.g., Amazon.com's "one-click" patent).

Further, the EPO requires that the software actually run and perform the specified process; the USPTO will grant software patents on concepts and theories not yet implemented. The JPO (Japanese Patent Office) will grant patents when the invention involves the concrete use of hardware resources. Software was not originally patentable but is increasingly being permitted, and the data being manipulated can be nontechnical (such as financial data).

Even though the major patenting authorities vary in what software and business-method patents they will grant, there's nothing to prevent applicants from filing forms with unpatentable software claims in the EPO and JPO. When they are published as "A" documents they become prior art to allowable U.S. software and business-method claims.

Unquestionably, business-method patents are controversial. The different countries' patent laws are in flux and are being widely debated. They will most likely continue to change. Thomson Derwent will not speculate on what the laws will evolve into, but they see a definite need for resources to help users find the prior art—both patents and publications—given the sheer volume of patents currently being published and issued in the area.

Derwent Web of Software includes the following resources:

• Derwent created a subset of the DWPI (Derwent World Patents Index) of patents in U.S. class 705 and Derwent's own software classes, plus additional patents produced by keyword searching—about 570,000 patents in all. The references include Derwent's abstracts and citation information from the Derwent Patent Citation Index. They also include all Derwent indexing, which amounts just to Derwent classes and some electrical Manual Codes, plus International Patent Classes. 

• Records from 2,000 journals included in the ISI Web of Science (Science Citation Index and related indexes) 

• Records from another 1,000 journals and trade publications from Gale InfoTrac as a source of news, business, and less technical journal literature 

• Several hundred outside Web sites selected by an ISI editorial team and peer-reviewed for relevance and credibility. According to Holtzman, these Web sites contain information on inventions and scientific advances that is not available elsewhere.

Derwent Web of Software will be a subscription product. Thomson Derwent expects to negotiate with individual customers, but in general the price will be based on both the total number of users and number of simultaneous users and will most likely be a five-figure annual fee.

Note that all these components—the Derwent World Patents Index, Science Citation Index, the Gale databases, and of course countless Internet Web sites—exist separately and can be searched separately on a pay-as-you-go basis. In fact, the Derwent records with their indexing in the Web of Software are available to nonsubscribers. Derwent claims that the value added is the ability to search all the components with one search strategy, as well as the elimination of a certain amount of noise (since all the records included are by definition related to software or business methods).

However, Derwent does not plan to develop any separate indexing for these patents. Searching is pretty much by abstract text and patent classifications. Derwent's press release refers to the "comprehensive content" of the database but (wisely) does not promise comprehensive searches. In actuality, the nomenclature used in software patents is probably even less consistent and predictable than it is in older, more mature technologies. And it has changed over the years. What used to be called "statistical pattern recognition" is now "data mining." So searchers need to know not only how the technology is described now, but also how it was described in the past. And that's for the references that actually mention product names—many of them include just generic descriptions in obscure "patentese."

Another issue is where good prior art can appear. The Derwent Web of Software beta-test evaluation form (at http://www.derwent.com/webofsoftware/beta/summary_feedback.html) lists a number of additional document forms and asks the beta testers to recommend which ones would be useful. The options listed include conference proceedings and white papers, patent and other litigation records, full text of patents, and software-related copyright and trademark information. But much of what might contain unique prior art (such as old software user manuals, product advertisements, and brochures) is for the most part not available in machine-readable form, so adding it would involve probably more labor than Thomson Derwent would be willing to commit to.

Derwent is targeting companies above a certain minimum size (rather than individual users), but to some extent it does plan to sell the Web of Software to law firms, consultants, and patent information brokers to use for nonsubscribing clients. Derwent will negotiate limitations on redistributing its records based on clients' needs. Hopefully this is a first step toward Derwent's coming up with ways of providing some nonsubscriber access to all its products in order to take a little of the financial burden off its subscribers.