A Recent Seminar on Examination Guidelines on Computer Software-related Inventions

On July 28, 2021, Saint Island team attended a seminar organized by the Taiwan Patent Attorneys Association (TWPAA) regarding the recent amendments to the Patent Examination Guidelines on computer software-related (CSR) patent applications. The TWPAA invited a senior patent examiner from the Taiwan Intellectual Property Office (TIPO) to present and discuss the new revisions along with the examples introduced into the Guidelines to shed light on how the new examination should be applied. For a detailed introduction of the revised guidelines, please refer to our post here.

The seminar began with an introduction of the recent trends in patent examination of the IP5 (USPTO, EPO, CNIPA, KIPO, and JPO) that motivated the TIPO to update the CSR Examination Guidelines on new emerging technologies (NET) relating to CSR inventions. In particular, the TIPO was largely inspired by an NET/AI Roadmap discussed during an IP5 Heads Industry Meeting in June, 2021 that pinpoints the AI-powered initiatives a patent system can adopt, including data analysis and forcasting, AI-assisted patent classification, and information exchanges.

As reported in our recent post (see here), Section 3 of the CSR Examination Guidelines addresses the definition of invention in terms of CSR applications, and a clearer test for this purpose is introduced with the revisions. Additionally, the seminar touched on the issue of “mixed inventions,” i.e. inventions that involve elements that are patentable and elements that are unpatentable (e.g. use of mathematical equations). How are these inventions examined? To begin with, the claim text should be considered as a whole in the test of whether the mixed invention fulfills the definition of invention. Then, the examiner should identify the nature of the subject matter of the claimed invention and put it to the test of whether it is clearly compatible or incompatible with the definition of invention. Note there are three instead of two possible answers: clearly compatible, clearly incompatible, and no clear conclusion available. If the answer is the third one, the next question of the test comes into play: whether the software-based information processing is concretely realized by using hardware resources (Fig. 1).

Fig. 1 Decision Flowchart of Eligibility of Software-Related Inventions

The CSR Examination Guidelines further recommend that examiners assess an invention’s subject matter eligibility in conjunction with, rather than ahead of, other examinations such as inventive step to optimize the examination process. For instance, if an invention fails the test of inventive step (which may be easier to determine), questions regarding its compatibility with the definition of invention is no longer relevant.

In the seminar, there were questions raised regarding the examination of inventions addressing software-based information processing based on hardware resources—do they fulfill the definition of invention? Our take from the seminar is that, to make a patent-eligible claim, the claim drafter needs to specify the means by which the software-based information processing is concretely realized by using hardware resources. Merely disclosing the input (including data pre-processing) and the output (such as data visualization) of the process is not enough to render the invention patent eligible.

Regarding the slightly modified rules on enablement and sufficiency of disclosure, which are found in Section 2 of the Guidelines, the TIPO advised the seminar attendees of the necessity to include sufficient embodiments in a specification. If an applicant claims a broad scope while the specification only contains few embodiments that are not representative of the broad terminology used in the claim language, then the claim may be deemed not supported by the specification, especially when a person of ordinary skill in the art cannot infer from the embodiments the much broader claim text.

For example, case study 2 of Section 2.2.4 of the CSR Examination Guidelines mentions a machine learning model that claimed to use “facial features” as part of an input to predict a person’s weight, but the specification only discloses one “facial feature,” namely the angles between the cheek and the chin. However, no other “facial features” are disclosed and, as a skilled artisan may find, facial features are not unequivocally relevant to the claimed model or any model asserted to be able to predict human weight.  Therefore, the claim is considered not supported by the specification.

Section 4.2 of the Guidelines outlines factors to be considered in the examination of inventive step. “Advantageous effect,” one of the factors that affirm inventive step, was featured in the seminar. Our take from the seminar is that simply automating a manual process with computers to improve speed or accuracy might not score in this factor, whereas an improvement in algorithm could be considered an advantageous effect meriting inventive step. For more information regarding the amendments in this aspect, please refer to our post here.