KNOWLEDGE ELICITATION FOR AUTHORING PATENT CLAIMS

To safeguard an invention, the inventor must apply for a patent. The f irst step is to author a patent claim. The claim is the focal point of a patent disclosure and the actual subject of legal protection. Claim s contain crucial information about the invention and must be formulat ed according to precise syntactic, lexical, and stylistic rules, as sp ecified in the guidelines pioneered by the German Patent Office and co mmonly accepted in the US and other countries. To successfully author a patent claim, you need two distinct types of expert knowledge: knowl edge about the sublanguage of patents as legal documents and knowledge about the invention's technology. This is why inventors (who possess the technical knowledge) need the services of lawyers or patent expert s (who possess the legal knowledge) to author a claim. To automate and thereby simplify the process of patent authoring, a system must elici t both kinds of knowledge.Legal knowledge essentially manifests itself in the constraints on and preferences concerning the lexical and gram matical language elements in a patent disclosure. This type of knowled ge applies to all patents regardless of domain and need not be elicite d more than once. Technical knowledge and the language of its descript ion are much more varied; thus, its elicitation should be conducted se parately for each patent document. Authoring patent disclosures is a c omplex task well suited for human-computer interaction, and patent cla im composition is the most difficult part. Even for experts, claim ana lysis and synthesis is time-consuming. In hopes of alleviating this si tuation, we developed an experimental system for semiautomatic authori ng of patent claims. The first implementation of our workstation, desi gned for both inventors and patent experts, is devoted to patent claim composition.This article focuses on interactively eliciting technical knowledge from inventors. These techniques let us readily acquire kno wledge about different domains. When they are coupled with text genera tion, we can ultimately simplify text composition tasks faced by all s orts of professionals.