The European Patent Office refused to grant a software patent for an expert system for assessing patents. Here are the practical takeaways from the decision T 1817/14 (Expert system/SCHINDLER) of Technical Board of Appeal 3.5.06:
In a nutshell, the invention proposes to obtain from a patent (or patent application) and any prior art document elements of their respective technical teachings. These may be specified informally. The elements and their “relations”, expressing anticipation and contradiction between elements or sets of elements, are arranged in what is called an ANC matrix (“anticipates/non-ants/contradicts”). The information in this matrix can be queried by and is then displayed to the user.
The claimed method has two phases. The first phase leads to the creation of the ANC matrix which, in the second phase, is used to “automatically and instantly” produce responses to user queries. The major part of the first phase is done by the user. Especially the generation of the ANC is under user control. Only the processing of user queries is meant to be automated.
Claim 1 (main request)
Is it patentable?
First of all, the claims of all the requests were considered to lack clarity. Nevertheless, the decision for this case is ultimately based on inventive step.
The board concludes that the major part of claim 1 (of all requests) is a modelling procedure during which the user considers the items in the domain of interest (comprising, specifically, the patent/patent application and the documents being compared, the laws of nature and the items of a national patent system), extracts their relevant properties (elements, facts, relations), and “compiles” them “into” a formal language.
The board considers this procedure of information modelling to be an intellectual activity (effectively a method for performing mental acts which does not, per se, contribute to the technical character of an invention). Accordingly, a technical contribution of the present invention could only lie in the way in which the generation and use of the model are implemented.
The appellant essentially argued that particular features of the ANC data structure had to be considered to be technical. It stressed in particular that the ANC had to reflect the analysis of documents in terms of two different levels of granularity (“elements” and “fundamental facts of these elements”) and that it contained novel fields (e.g. “anticipates/not-anticipates-and-not-contradicts/contradicts” as claimed).
However, the appellant was unable to convince the Board that the modelling steps caused any technical effect. When, however, the modelling steps are assumed to be taken as an aim to be achieved in a non-technical field, the form of the ANC is determined by the model and thus obvious.
In addition, the feature that users may query the “items” in the ANC and the method replies “automatically and instantly by displaying to the user this item’s information and all its such relations to other items” does not go beyond the statement that the information in the ANC may be accessed by user queries, as is known from prior-art database systems.
The board concludes that claim 1 of all requests lacks inventive step in view of common knowledge, as an “obvious way of providing computer support to an essentially non-technical method“.
You can read the whole decision here: T 1817/14 (Expert system/SCHINDLER) of 4.7.2017